At @VeedaAI , we’re building for a future where robots learn by interacting with the world, at scale -- in simulated reality.
Getting there means bringing together the brightest minds to tackle some of the hardest problems in Physical AI.
Come build what’s next with us:
Sanja Fidler
Sanja Fidler is a Slovenian-Canadian computer scientist. She is an associate professor at the University of Toronto Mississauga and the chief executive officer of Veeda Innovation Inc. Previously, she was vice-president of artificial intelligence (AI) research at Nvidia from 2022 to 2026, leading Nvidia's Spatial Intelligence Lab. She is also a co-founder of the Vector Institute and CIFAR AI Chair. Her research is in the areas of computer vision and artificial intelligence. More on Wikipedia
-
Sanja Fidler
@FidlerSanja
-
Sanja Fidler reposted
Huan Ling
@HuanLing6
Today, I’m incredibly excited to introduce Veeda AI, which I co-founded with @FidlerSanja and @ZGojcic .
Throughout my career, I’ve been fascinated by one question: how can visual generative models impact the physical world?
My journey has taken me from images to video to world foundation models.
Now, we’re building toward the next frontier: turning these generated worlds into virtual interactive playgrounds where physical intelligence can perceive, reason, act, and ultimately learn within them at scale.
That’s our ambition at Veeda and I’m incredibly excited to build this alongside Sanja, Zan, and our amazing team.
We’re hiring. Come build it with us.
Sanja Fidler
@FidlerSanja
I’m incredibly excited to introduce @VeedaAI , which I co-founded with my incredible longtime collaborators @ZGojcic and @HuanLing6 .
We started Veeda because we believe robotics will reshape the world—changing how we move people and goods, how we manufacture and build, and how we operate in the physical world.
We also firmly believe that the scaling moment for Physical AI will come from robots learning through interaction with the world.
Just like humans learn from interaction with their environments by trying, failing and trying again, until we succeed. And just like the capabilities of LLMs were truly unlocked when they started training in interactive environments and not only on vast amounts of internet text.
For Physical AI, the central challenge is making this kind of interactive learning possible at scale. The real world is simply not a practical training ground for robots to learn through trial and error. It is unsafe, expensive, and time-consuming. To scale interactive learning, robots will need to learn in simulated reality.
At Veeda, our sole mission is to build simulated reality for Physical AI. Our conviction is that this will become the critical infrastructure layer for all areas of robotics.
As envisioned in pop culture, this means building “the Matrix” for Physical AI, where, through a virtual embodiment, robot intelligence can interact with an entirely virtual world in a million possible ways, learning from its own mistakes.
We believe that World Models are the foundational technology that can make this possible. These generative models learn from massive amounts of sensor and physical-world data to reach the quality, diversity, and physical realism of the simulated reality that robots require.
I could not be more excited to take on this challenge and build this transformative technology alongside an extraordinary team of engineers and researchers who have spent years working on some of the hardest problems across simulation, generative AI, 3D, robotics, and embodied intelligence.
Together with our investors at @khoslaventures and @radicalvcfund , we intend to push the frontier of this technology and build the foundational infrastructure for interactive robotics learning and evaluation at scale.
Come build this future with us.
And to roboticists, help us understand what you need, we’re building this for you.
veeda.ai -
Sanja Fidler
@FidlerSanja
I’m incredibly excited to introduce @VeedaAI , which I co-founded with my incredible longtime collaborators @ZGojcic and @HuanLing6 .
We started Veeda because we believe robotics will reshape the world—changing how we move people and goods, how we manufacture and build, and how we operate in the physical world.
We also firmly believe that the scaling moment for Physical AI will come from robots learning through interaction with the world.
Just like humans learn from interaction with their environments by trying, failing and trying again, until we succeed. And just like the capabilities of LLMs were truly unlocked when they started training in interactive environments and not only on vast amounts of internet text.
For Physical AI, the central challenge is making this kind of interactive learning possible at scale. The real world is simply not a practical training ground for robots to learn through trial and error. It is unsafe, expensive, and time-consuming. To scale interactive learning, robots will need to learn in simulated reality.
At Veeda, our sole mission is to build simulated reality for Physical AI. Our conviction is that this will become the critical infrastructure layer for all areas of robotics.
As envisioned in pop culture, this means building “the Matrix” for Physical AI, where, through a virtual embodiment, robot intelligence can interact with an entirely virtual world in a million possible ways, learning from its own mistakes.
We believe that World Models are the foundational technology that can make this possible. These generative models learn from massive amounts of sensor and physical-world data to reach the quality, diversity, and physical realism of the simulated reality that robots require.
I could not be more excited to take on this challenge and build this transformative technology alongside an extraordinary team of engineers and researchers who have spent years working on some of the hardest problems across simulation, generative AI, 3D, robotics, and embodied intelligence.
Together with our investors at @khoslaventures and @radicalvcfund , we intend to push the frontier of this technology and build the foundational infrastructure for interactive robotics learning and evaluation at scale.
Come build this future with us.
And to roboticists, help us understand what you need, we’re building this for you.
veeda.ai -
Sanja Fidler reposted
Jiahui Huang
@huangjh_hjh
Check out #instantnurec, which performs feed-forward 3D Gaussian reconstruction for driving scene simulation within seconds.
We have now made this into the official NVIDIA #omniverse NuRec product. Code and models are available.
research.nvidia.com/labs/sil/projects/instant-nurec/ -
Sanja Fidler reposted
Ruilong Li
@ruilong_li
Actually, both of these two NVIDIA live demos at CVPR are powered by flashdreams!
Ruilong Li
@ruilong_li
World models are moving beyond offline generation towards interactive, real-time experiences.
Introducing ⚡FlashDreams⚡: an open-source high-performance inference and serving library built for autoregressive world models:
🔥 Up to 3.10× faster LingBot-World inference
🔥 Up to 2.12× faster Self-Forcing inference
🔥 Up to 1.40× faster Wan2.1 inference
🔥 8 integrated models
🔥 Multi-GPU, streaming, low-latency serving
🔥 Agentic skills that teach you how to use it
FlashDreams is designed for a new generation of AI systems that continuously evolve over time while responding to user interactions. It powers applications across robotics, autonomous vehicle simulation, gaming, and virtual worlds.
Github: github.com/NVIDIA/flashdreams
Docs: nvidia.github.io/flashdreams
Research page: research.nvidia.com/labs/sil/projects/flashdreams/
Join the #flashdreams Discord channel at discord.gg/yTdHDqFP
FlashDreams is also the runtime backbone behind NVIDIA OmniDreams (github.com/nv-tlabs/omni-dreams)
1/n
#ai #worldmodels #fastinference #physicalai #opensource #nvidia -
Sanja Fidler reposted
Zian Wang
@zianwang97
Following recent World-Action Model results in robotics, the same ~2B OmniDreams single-view backbone can be fine-tuned into a driving policy.
In preliminary closed-loop results, it reduces collision from 6.9% to 4.2% when compared with Alpamayo 1.5, while having roughly 5x fewer parameters. -
Sanja Fidler reposted
Michał Tyszkiewicz
@jatentaki
Alpasim now lets you test your AV policies with an interactive world model!
-
Sanja Fidler reposted
Despoina Paschalidou
@paschalidoud_1
It’s been a while since I posted here, but I’m very excited to share what our team at @nvidia has been building over the past year!
After a year of active development, we’re getting ready to release SIL-Wheel to the world: a one-stop shop platform for data-centric workflows in large-scale video model training.
Built by researchers, for researchers, SIL-Wheel brings together search, curation, annotation, evaluation, and analysis for large video datasets in one centralized framework.
Want a sneak peek before the official release? Come by the NeXD26 Workshop @CVPR tomorrow at 10:30!🚀 -
Sanja Fidler
@FidlerSanja
Real time world model NVIDIA OmniDreams now open sourced!
If you are at CVPR, we invite you to also check out a live demo you can try out at the NVIDIA booth.
Zian Wang
@zianwang97
🚀 What if physical AI policies could interact with generated worlds in real time?
Introducing OmniDreams, a generative world model for closed-loop autonomous vehicle simulation.
Tech report, code, models, and data samples are available now.
Project: research.nvidia.com/labs/sil/projects/omnidreams-blog/
Code: github.com/nv-tlabs/omni-dreams
Model: huggingface.co/collections/nvidia/nvidia-omnidreams
Join the #omnidreams discord channel: discord.gg/bsjzh4uZ -
Sanja Fidler reposted
Zian Wang
@zianwang97
🚀 What if physical AI policies could interact with generated worlds in real time?
Introducing OmniDreams, a generative world model for closed-loop autonomous vehicle simulation.
Tech report, code, models, and data samples are available now.
Project: research.nvidia.com/labs/sil/projects/omnidreams-blog/
Code: github.com/nv-tlabs/omni-dreams
Model: huggingface.co/collections/nvidia/nvidia-omnidreams
Join the #omnidreams discord channel: discord.gg/bsjzh4uZ -
Sanja Fidler
@FidlerSanja
Super excited to open source Flashdreams, acceleration library for world models! No restrictions on usage, Apache 2 license.
World models are the next frontier for video generation and paramount to Physical AI.
Flashdreams yields out of the box speed up across world models
Ruilong Li
@ruilong_li
World models are moving beyond offline generation towards interactive, real-time experiences.
Introducing ⚡FlashDreams⚡: an open-source high-performance inference and serving library built for autoregressive world models:
🔥 Up to 3.10× faster LingBot-World inference
🔥 Up to 2.12× faster Self-Forcing inference
🔥 Up to 1.40× faster Wan2.1 inference
🔥 8 integrated models
🔥 Multi-GPU, streaming, low-latency serving
🔥 Agentic skills that teach you how to use it
FlashDreams is designed for a new generation of AI systems that continuously evolve over time while responding to user interactions. It powers applications across robotics, autonomous vehicle simulation, gaming, and virtual worlds.
Github: github.com/NVIDIA/flashdreams
Docs: nvidia.github.io/flashdreams
Research page: research.nvidia.com/labs/sil/projects/flashdreams/
Join the #flashdreams Discord channel at discord.gg/yTdHDqFP
FlashDreams is also the runtime backbone behind NVIDIA OmniDreams (github.com/nv-tlabs/omni-dreams)
1/n
#ai #worldmodels #fastinference #physicalai #opensource #nvidia -
Sanja Fidler reposted
Ruilong Li
@ruilong_li
World models are moving beyond offline generation towards interactive, real-time experiences.
Introducing ⚡FlashDreams⚡: an open-source high-performance inference and serving library built for autoregressive world models:
🔥 Up to 3.10× faster LingBot-World inference
🔥 Up to 2.12× faster Self-Forcing inference
🔥 Up to 1.40× faster Wan2.1 inference
🔥 8 integrated models
🔥 Multi-GPU, streaming, low-latency serving
🔥 Agentic skills that teach you how to use it
FlashDreams is designed for a new generation of AI systems that continuously evolve over time while responding to user interactions. It powers applications across robotics, autonomous vehicle simulation, gaming, and virtual worlds.
Github: github.com/NVIDIA/flashdreams
Docs: nvidia.github.io/flashdreams
Research page: research.nvidia.com/labs/sil/projects/flashdreams/
Join the #flashdreams Discord channel at discord.gg/yTdHDqFP
FlashDreams is also the runtime backbone behind NVIDIA OmniDreams (github.com/nv-tlabs/omni-dreams)
1/n
#ai #worldmodels #fastinference #physicalai #opensource #nvidia -
Sanja Fidler reposted
Xuanchi Ren
@xuanchi13
The latent-vs-pixel debate misses the point.
GPT Image 2 shows what users notice: pixel-level fidelity.
Latent models show what scales: compact semantic structure.
We connect them by replacing VAE/RAE decoders with a Pixel Diffusion Decoder.
Code and Model available: research.nvidia.com/labs/sil/projects/pid/
🧵(1/N)This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Jonathan Lorraine
@jonLorraine9
Our paper has been accepted to #icml2026 as a Spotlight (top 2.2%) 🥳
We present MOTIVE, a scalable, motion-centric data-attribution framework for video generation that identifies which training clips improve or degrade temporal dynamics, enabling curation and more.
Xindi Wu
@cindy_x_wu
New #nvidia Paper
We introduce Motive, a motion-centric, gradient-based data attribution method that traces which training videos help or hurt video generation.
By isolating temporal dynamics from static appearance, Motive identifies which training videos shape motion in video generation.
🔗 research.nvidia.com/labs/sil/projects/MOTIVE/
1/10 -
Sanja Fidler reposted
Xindi Wu
@cindy_x_wu
New #nvidia Paper
We introduce Motive, a motion-centric, gradient-based data attribution method that traces which training videos help or hurt video generation.
By isolating temporal dynamics from static appearance, Motive identifies which training videos shape motion in video generation.
🔗 research.nvidia.com/labs/sil/projects/MOTIVE/
1/10 -
Sanja Fidler reposted
Kangxue Yin
@kangxue_yin
🚀We just released Asset Harvester, an image-to-3D model and end-to-end pipeline that extracts real object assets from autonomous driving videos!
🌐 Website: research.nvidia.com/labs/sil/projects/asset-harvester/
💻 Code: github.com/nvidia/asset-harvester/
[1/5]
#assetharvester #avsimulation #worldmodel #autonomousdriving -
Sanja Fidler reposted
Kai He
@Kai__He
We open-sourced the code and model for UniRelight! 🎉
Given an input video and a target lighting configuration, our method jointly predicts a relit video and its corresponding albedo.
Code: github.com/nv-tlabs/UniRelight
Model: huggingface.co/nvidia/UniRelight
Kai He
@Kai__He
🚀 Introducing UniRelight, a general-purpose relighting framework powered by video diffusion models.
🌟UniRelight jointly models the distribution of scene intrinsics and illumination, enabling high-quality relighting and intrinsic decomposition from a single image or video.This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Janick Martinez Esturo
@jmartinezesturo
A canonical open-source data platform for multi-sensor neural 3D reconstruction is here.
Introducing NVIDIA NCore — a unified data format and APIs for cameras, lidars, radars, poses, calibrations, and labels, built for AV, robotics, and physical AI.
🔗research.nvidia.com/labs/sil/projects/ncore -
Sanja Fidler reposted
Michał Tyszkiewicz
@jatentaki
Feed-forward 3D reconstruction should not be limited to predicting one Gaussian per pixel.
We introduce TokenGS, which uses learnable tokens to decouple the 3D Gaussian prediction from the image resolution and the number of input views.
#cvpr2026highlight
[1/6] -
Sanja Fidler reposted
Xuanchi Ren
@xuanchi13
We scaled up Lyra to generate explorable 3D worlds! 🚀
Introducing Lyra 2.0 — turning a single image into a 3D world you can walk through, look back, and even drop a robot into 🤖
Code and Model available today!
🌐 Website: research.nvidia.com/labs/sil/projects/lyra2/
(1/N)This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Today, we released Lyra 2.0, a framework for generating persistent, explorable 3D worlds at scale, from NVIDIA Research.
Generating large-scale, complex environments is difficult for AI models. Current models often “forget” what spaces look like and lose track of movement over time, causing objects to shift, blur, or appear inconsistent. This prevents them from creating the reliable 3D environments required for downstream simulations. Lyra 2.0 solves these issues by:
✅ Maintaining per-frame 3D geometry to retrieve past frames and establish spatial correspondences
✅ Using self-augmented training to correct its own temporal drifting.
Lyra 2.0 turns an image into a 3D world you can walk through, look back, and drop a robot into for real-time rendering, simulation, and immersive applications.
➡️ Learn more: research.nvidia.com/labs/sil/projects/lyra2/
📄 Read the paper: arxiv.org/abs/2604.13036 -
Sanja Fidler reposted
NVIDIA DRIVE
@NVIDIADRIVE
This Cosmos-based simulation demo caught Jensen's attention.
See what it's all about at the NVIDIA booth.
#nvidiagtc -
Sanja Fidler reposted
Chen Tessler @ CoRL
@ChenTessler
Animation 🤝 Robotics
ProtoMotions GTC 2026 release — bridging the gap between digital humans and real humanoid robots.
Train in simulation. Deploy on hardware. One framework, one codebase.
nvlabs.github.io/ProtoMotions -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
The most mind-blowing part: the steering wheel demo runs entirely locally on a DGX station announced by Jensen today. No cloud. Just pure on-device intelligence.
Come test it out at the GTC booth!
Zan Gojcic
@ZGojcic
AlpaDreams is integrated into AlpaSim (github.com/NVlabs/alpasim) and can be controlled either by a policy model such as Alpamayo 1 or directly by a human using a steering wheel—both in real time.
-
Sanja Fidler reposted
Davis Rempe
@davrempe
Need high-quality motion for humanoid robots or digital humans?
Meet Kimodo: our new diffusion model trained on 700 hours of optical mocap data for easy, controllable, and high-fidelity motion generation. @NVIDIAAI
research.nvidia.com/labs/sil/projects/kimodo/ -
Sanja Fidler reposted
Zian Wang
@zianwang97
At #gtc26 we are showing AlpaDreams — generative, interactive closed-loop simulation for autonomous driving.
World models generate environments in real time, policies or drivers interact, and the simulator updates the world.
If you are at GTC, come by the booth and try it out! -
Sanja Fidler reposted
Jonathan Lorraine
@jonLorraine9
🌎🚗 Introducing AlpaDreams from our group at NVIDIA's Spatial Intelligence Lab: a real-time generative world model for closed-loop AV simulation.
See more on the project page: research.nvidia.com/labs/sil/projects/alpadreams/ -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
A new generation in AV simulation is here!
We are announcing AlpaDreams, a real time interactive generative world model for AV simualtion! Just a year ago it took minutes to generate a few seconds of video, today it is real time and interactive!
research.nvidia.com/labs/sil/projects/alpadreams/ -
Sanja Fidler reposted
Haotian Ye
@haotian_yeee
🤔Want a principled way to RL your diffusion model?
Check Data-regularized Reinforcement Learning (DDRL)! Post-train @nvidia #cosmos World Foundation models with a million GPU hours! 🤯
Novel formulation ➡️ Theoretically integrates SFT into RL ➡️ Robust to Reward Hacking 🛑
Details: research.nvidia.com/labs/dir/ddrl/
#ddrl #diffusion #rl #nvidia #cosmos -
Sanja Fidler reposted
Chen Tessler @ CoRL
@ChenTessler
At @nvidia, we built ProtoMotions to help us, and researchers world-wide, innovate quickly without compromising on applicability.
We're proud to announce ProtoMotions3 -- our biggest release yet!
🧵👇 -
Sanja Fidler reposted
AK
@_akhaliq
World Simulation with Video Foundation Models for Physical AI
-
Sanja Fidler reposted
AK
@_akhaliq
Nvidia just released ChronoEdit-14B on Hugging Face
enables physics-aware image editing and action-conditioned world simulation through temporal reasoning. It distills priors from a 14B-parameter pretrained video generative model and separates inference into (i) a video reasoning stage for latent trajectory denoising, and (ii) an in-context editing stage for pruning trajectory tokens. ChronoEdit-14B was developed by NVIDIA as part of the ChronoEdit family of multimodal foundation models -
Sanja Fidler reposted
Xiuming Zhang
@xiuming_zhang
📢 NVIDIA's Spatial Intelligence Lab (research.nvidia.com/labs/sil/) is hiring 2026 research interns for video/content generation, generative world simulation, RL, 4D perception, neural reconstruction, geometry processing, and physics simulation! Interested? Ping me here or via email.
-
Sanja Fidler reposted
Huan Ling
@HuanLing6
🕹️We are excited to introduce "ChronoEdit: Towards Temporal Reasoning for Image Editing and World Simulation"
ChronoEdit reframes image editing as a video generation task to encourage temporal consistency. It leverages a temporal reasoning stage that denoises with “video reasoning tokens” to "reason" on physically plausible edits.
See the attached video for results.
Project Page: research.nvidia.com/labs/toronto-ai/chronoedit/
Arxiv: arxiv.org/abs/2510.04290
Code and model are coming. -
Sanja Fidler reposted
Jiahui Huang
@huangjh_hjh
In Lyra, we reconstruct dynamic 3D scenes directly from the video model's latent space, achieving real-time interaction and perfect 3D consistency!
Glad that ViPE and GEN3C are the key enabling technologies in this work.
Kudos to @sherwinbahmani @xuanchi13 @JunGao33210520!
Sherwin Bahmani
@sherwinbahmani
📢 Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation
Got only one or a few images and wondering if recovering the 3D environment is a reconstruction or generation problem? Why not do it with a generative reconstruction model!
We show that a camera-conditioned video diffusion model can be transformed into a generative reconstruction model that directly outputs a high-quality 3D Gaussian Splatting representation through self-distillation, without requiring real-world training data.
Check out our results in the video (wait for dynamic scenes in the second half!) :
Project Page: research.nvidia.com/labs/toronto-ai/lyra/
Code and Models: github.com/nv-tlabs/lyra
Paper: arxiv.org/abs/2509.19296This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Sherwin Bahmani
@sherwinbahmani
The generated 3D Gaussians can then be exported into into NVIDIA Isaac Sim based on 3DGUT to simulate humanoid robots in generated environments.
-
Sanja Fidler reposted
Sherwin Bahmani
@sherwinbahmani
📢 Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation
Got only one or a few images and wondering if recovering the 3D environment is a reconstruction or generation problem? Why not do it with a generative reconstruction model!
We show that a camera-conditioned video diffusion model can be transformed into a generative reconstruction model that directly outputs a high-quality 3D Gaussian Splatting representation through self-distillation, without requiring real-world training data.
Check out our results in the video (wait for dynamic scenes in the second half!) :
Project Page: research.nvidia.com/labs/toronto-ai/lyra/
Code and Models: github.com/nv-tlabs/lyra
Paper: arxiv.org/abs/2509.19296This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Ruofan Liang
@RfLiang
💡 Introducing LuxDiT: a diffusion transformer (DiT) that estimates realistic scene lighting from a single image or video.
It produces accurate HDR environment maps, addressing a long-standing challenge in computer vision.
🔗Paper: arxiv.org/abs/2509.03680 -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
🤳Selfies, dashcams, 360° -- ViPE solves one of spatial AI’s most difficult challenges by recovering camera motion from ANY video 🎥
We shared ViPE at #siggraph25 using a single GPU processing thousands of frames in just minutes while maintaining stability and consistency– providing up to 50% improvement over previous pose methods.
Try it here 🔗 nvda.ws/41AVayH -
Sanja Fidler reposted
Jonathan Stephens
@jonstephens85
The team at @NVIDIAAIDev really nailed it with ViPE!! This SLAM system generated pose, depth, point clouds, etc, in a few seconds! Now to get the output to work with GEN3C...
If you are interested in running this yourself, it took about 5 minutes to get up and running from scratch. github.com/nv-tlabs/vipeThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
3DGRT is now available also with Vulkan API as part of the Vulkan Gaussian Splatting:
github.com/nvpro-samples/vk_gaussian_splatting
Zan Gojcic
@ZGojcic
📢📢 3D Gaussian Ray Tracing: Fast Tracing of Particle Scenes
We are tracing 3D Gaussians, and we do it fast!
3DGRT natively supports distorted cameras, rolling-shutter, motion blur, depth of field, refraction, reflections, and shadows! #siggraphasia2024 @NVIDIAAI -
Sanja Fidler reposted
MrNeRF
@janusch_patas
ViPE: Video Pose Engine for 3D Geometric Perception
Contributions:
• A robust and efficient framework, ViPE, for estimating camera parameters and dense depth from diverse, in-the-wild videos.
• A system design that integrates the strengths of classical SLAM (efficiency, scalability) and learned models (robustness), with key improvements in efficiency, dynamic object handling, and depth quality over prior work.
• A large-scale dataset of annotated videos, created using ViPE, to facilitate future research in 3D computer vision. -
Sanja Fidler reposted
Jiahui Huang
@huangjh_hjh
[7/N] ViPE code is released at: github.com/nv-tlabs/vipe!
Huge thanks to the incredible team behind this @QunjieZhou @HesamRabeti Aleksandr @HuanLing6 @xuanchi13 @TianchangS @JunGao33210520 Dmitry @chenhsuanlin @jiawei6_ren Kevin @Joydeepb_robots @lealtaixe @FidlerSanja -
Sanja Fidler reposted
Jiahui Huang
@huangjh_hjh
[1/N] 🎥 We've made available a powerful spatial AI tool named ViPE: Video Pose Engine, to recover camera motion, intrinsics, and dense metric depth from casual videos!
Running at 3–5 FPS, ViPE handles cinematic shots, dashcams, and even 360° panoramas.
🔗 research.nvidia.com/labs/toronto-ai/vipe/ -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
📢📢 "Align Your Flow: Scaling Continuous-Time Flow Map Distillation"
New flow map framework for state-of-the-art few-step generation, w/ the amazing @amsabour and @FidlerSanja.
🔥 Project page: research.nvidia.com/labs/toronto-ai/AlignYourFlow/
📜 Paper: arxiv.org/abs/2506.14603
🧵Thread below...
(1/n) -
Sanja Fidler reposted
AK
@_akhaliq
Nvidia just announced Align Your Flow
Scaling Continuous-Time Flow Map Distillation -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
To wrap up the open-sourcing season, we’re excited to announce that DiffusionRenderer, based on the NVIDIA Cosmos world model, is now open sourced!
That means that the code of all three of our #cvpr25 oral papers is now available:
- 3DGUT
- DiffusionRenderer
- Diffix3D+
@CVPR
Zian Wang
@zianwang97
🚀 DiffusionRenderer is now open-source!
Check out the code and model: github.com/nv-tlabs/diffusion-renderer
We will present at #cvpr2025 this Sunday, June 15:
🗣️ Oral Session 6A: 1:00–2:30 PM CDT, Karl Dean Grand Ballroom
🖼️ Poster: 4:00–6:00 PM CDT, ExHall D (Poster #29)
Zian Wang
@zianwang97
🚀 Introducing DiffusionRenderer, a neural rendering engine powered by video diffusion models.
🎥 Estimates high-quality geometry and materials from videos, synthesizes photorealistic light transport, enables relighting and material editing with realistic shadows and reflectionsThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Sophia
@sopharicks
Excited to host @ZGojcic talk next week on @nvidia DiffusionRenderer, a new technique for neural rendering. It approximates how light behaves in the real world and can turn daytime scenes into night, sunny scenes into cloudy ones, and so on. It combines inverse and forward rendering into a unified neural engine that outperforms SOTA methods.
The talk is online. Use the link below to register and join the discussion with Zan lu.ma/l9yse44r -
Sanja Fidler reposted
Zian Wang
@zianwang97
🚀 We just open-sourced Cosmos DiffusionRenderer!
This major upgrade brings significantly improved video de-lighting and re-lighting—powered by NVIDIA Cosmos and enhanced data curation.
Released under Apache 2.0 and Open Model License. Try it out!
🔗 github.com/nv-tlabs/cosmos1-diffusion-rendererThis media couldn't be displayed. See it on the original post.
Zian Wang
@zianwang97
🚀 Introducing DiffusionRenderer, a neural rendering engine powered by video diffusion models.
🎥 Estimates high-quality geometry and materials from videos, synthesizes photorealistic light transport, enables relighting and material editing with realistic shadows and reflectionsThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Jay Wu
@jayzhangjiewu
🚀 Difix3D+ is now open-sourced!
Check out the code and try the demo: github.com/nv-tlabs/Difix3D
We're presenting at #cvpr2025 this Sunday, June 15 — come say hi!
🗣️ Oral: 1:00–1:15 PM CDT, Karl Dean Grand Ballroom
🖼️ Poster: 4:00–6:00 PM CDT, ExHall D (Poster #57)
Jay Wu
@jayzhangjiewu
Excited to share our #cvpr2025 paper: Difix3D+
Difix3D+ reimagines 3D reconstruction with single-step diffusion, distilling 2D generative priors for realistic novel view synthesis from large viewpoint shifts.
📄Paper: arxiv.org/abs/2503.01774
🌐Website: research.nvidia.com/labs/toronto-ai/difix3d -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
🚀 Just in time for the #cvpr rush: we’ve released the code for Difix3D+ — a Best Paper Award candidate!
🔧 Try out the code & demos: github.com/nv-tlabs/Difix3D
🎤 Oral (June 15): 1:00–1:15 PM CDT, Karl Dean Grand Ballroom
🖼️ Poster: 4:00–6:00 PM CDT, ExHall D (#57)
Join us! @CVPR
Jay Wu
@jayzhangjiewu
🚀 Difix3D+ is now open-sourced!
Check out the code and try the demo: github.com/nv-tlabs/Difix3D
We're presenting at #cvpr2025 this Sunday, June 15 — come say hi!
🗣️ Oral: 1:00–1:15 PM CDT, Karl Dean Grand Ballroom
🖼️ Poster: 4:00–6:00 PM CDT, ExHall D (Poster #57)
Jay Wu
@jayzhangjiewu
Excited to share our #cvpr2025 paper: Difix3D+
Difix3D+ reimagines 3D reconstruction with single-step diffusion, distilling 2D generative priors for realistic novel view synthesis from large viewpoint shifts.
📄Paper: arxiv.org/abs/2503.01774
🌐Website: research.nvidia.com/labs/toronto-ai/difix3d -
Sanja Fidler reposted
Huan Ling
@HuanLing6
We are excited to share Cosmos-Drive-Dreams 🚀
A bold new synthetic data generation (SDG) pipeline powered by world foundation models—designed to synthesize rich, challenging driving scenarios at scale.
Models, Code, Dataset, Tookit are released.
Website:This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Xuanchi Ren
@xuanchi13
GEN3C #cvpr2025 is now open-sourced @HuggingFace, with code released under Apache 2.0 and model weights under the NVIDIA Open Model License!
Code: github.com/nv-tlabs/GEN3C
HuggingFace: huggingface.co/nvidia/GEN3C-Cosmos-7B
Feel free to reach out if you want to chat or connect! #cvpr2025
AK
@_akhaliq
Nvidia just dropped GEN3C
3D-Informed World-Consistent Video Generation with Precise Camera ControlThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Radiance Fields
@RadianceFields
NVIDIA just open sourced GEN3C with an Apache 2.0 license at CVPR!
github.com/nv-tlabs/GEN3C
Xuanchi Ren
@xuanchi13
🚀Excited to introduce GEN3C #cvpr2025, a generative video model with an explicit 3D cache for precise camera control.
🎥It applies to multiple use cases, including single-view and sparse-view NVS🖼️ and challenging settings like monocular dynamic NVS and driving simulation🚗.
Project page: research.nvidia.com/labs/toronto-ai/GEN3C/ -
Sanja Fidler reposted
Jun Gao
@JunGao33210520
This year, we have 3 papers in CVPR, discussing the connection between 3D and video models:
GEN3C [Highlight] 3D grounding for video model
DiffusionRenderer [Oral] Taming video models for rendering and inverse rendering
Diffix3D+ [Oral] Enhancing Nerf/3DGS w/ diffusion models -
Sanja Fidler reposted
Tianchang Shen
@TianchangS
📢 GEN3C is now open-sourced, with code released under Apache 2.0 and model weights under the NVIDIA Open Model License!
🚀 Along with it, we're releasing a GUI tool that lets you specify your desired video trajectory in 3D — come play with it and generate your own!
The released model is built on COSMOS-7B, bringing better visual quality, stronger generalization, and longer, more coherent video generation than SVD-based GEN3C.
Heading to #cvpr2025? Stop by our poster — ExHall D, #65, Friday June 13, 2–4 p.m. PDT. Come say hi! 👋
Code: github.com/nv-tlabs/GEN3C
Model: huggingface.co/nvidia/GEN3C-Cosmos-7BThis media couldn't be displayed. See it on the original post.
Xuanchi Ren
@xuanchi13
🚀Excited to introduce GEN3C #cvpr2025, a generative video model with an explicit 3D cache for precise camera control.
🎥It applies to multiple use cases, including single-view and sparse-view NVS🖼️ and challenging settings like monocular dynamic NVS and driving simulation🚗.
Project page: research.nvidia.com/labs/toronto-ai/GEN3C/ -
Sanja Fidler reposted
Chih-Hao Lin
@ZhiHaoLin16
What if you could control the weather in any video — just like applying a filter?
Meet WeatherWeaver, a video model for controllable synthesis and removal of diverse weather effects — such as 🌧️ rain, ☃️ snow, 🌁 fog, and ☁️ clouds — for any input video. -
Sanja Fidler reposted
Tianchang Shen
@TianchangS
FlexiCubes is now under Apache 2.0! 🎉
We've been excited to see FlexiCubes extracting high-quality meshes across the community in projects like TRELLIS and TripoSF --- now it's available with a more permissive license.
Let's keep building. 💙
👉 Flexicubes is in NVIDIA Kaolin: github.com/NVIDIAGameWorks/kaolin
👉 Or a standalone repo github.com/nv-tlabs/FlexiCubes -
Sanja Fidler reposted
Erwin Coumans 🇺🇦 ✈️ CoRL 2026
@erwincoumans
So the PhysX CUDA kernel source code is now open source!!!
github.com/NVIDIA-Omniverse/PhysX/blob/main/physx/CHANGEL… and github.com/NVIDIA-Omniverse/PhysX/tree/main/physx/source -
Sanja Fidler reposted
Devi Parikh
@deviparikh
One year ago, @abhshkdz and I left Meta to start Yutori.
Ten months ago, @DhruvBatraDB joined us :)
Nine months ago, we crystallized our vision.
Two months ago, we released a sneak peak into what we’ve been building.
Today, can’t be more excited to fully unveil @yutori_ai’s mission, ambitious vision, world-class team, and stellar backers!
Join our waitlist for early access to our product in the coming days and weeks! 👇🏼 -
Sanja Fidler reposted
Miles Macklin
@milesmacklin
Big news from Jensen at GTC: NVIDIA Warp is now open source under Apache 2.0! We’re also building Newton, a new physics engine for robotics with DeepMind and Disney—based on Warp and fully open-source.
Christopher Crouzet
@christophercrzt
Btw... Warp just transitioned to an Apache 2.0 license! 👀 developer.nvidia.com/warp-python
-
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Spatial AI is increasingly important, and the newest papers from #nvidiaresearch, 3DGRT and 3DGUT, represent significant advancements in enabling researchers and developers to explore and innovate with 3D Gaussian Splatting techniques.
💎 3DGRT (Gaussian Ray Tracing) ➡️ research.nvidia.com/labs/toronto-ai/3DGRT/
- Enables secondary lighting effects such as refractions, shadows, depth of field, and mirrors.
- Provides a foundation for further research in ray tracing-based rendering of Gaussian representations.
- Achieves real-time rendering performance on modern GPUs.
✨ 3DGUT (Gaussain Unscented Transform) ➡️ research.nvidia.com/labs/toronto-ai/3DGUT/
- Support for distorted camera models and time-dependent effects like rolling shutter.
- Enables phenomena such as reflections and refractions while retaining real-time efficiency.
📥 Code: github.com/nv-tlabs/3dgrut/ -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
📣 We have released the source code of 3DGRT and 3DGUT in a common 3DGRUT repo under the Apache license!
Go try it out and play with our playground app!
NVIDIA AI Developer
@NVIDIAAIDev
Spatial AI is increasingly important, and the newest papers from #nvidiaresearch, 3DGRT and 3DGUT, represent significant advancements in enabling researchers and developers to explore and innovate with 3D Gaussian Splatting techniques.
💎 3DGRT (Gaussian Ray Tracing) ➡️ research.nvidia.com/labs/toronto-ai/3DGRT/
- Enables secondary lighting effects such as refractions, shadows, depth of field, and mirrors.
- Provides a foundation for further research in ray tracing-based rendering of Gaussian representations.
- Achieves real-time rendering performance on modern GPUs.
✨ 3DGUT (Gaussain Unscented Transform) ➡️ research.nvidia.com/labs/toronto-ai/3DGUT/
- Support for distorted camera models and time-dependent effects like rolling shutter.
- Enables phenomena such as reflections and refractions while retaining real-time efficiency.
📥 Code: github.com/nv-tlabs/3dgrut/ -
Sanja Fidler reposted
AK
@_akhaliq
Nvidia just released Cosmos-Transfer1 on Hugging Face
Conditional World Generation with Adaptive Multimodal ControlThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Huan Ling
@HuanLing6
Thank you AK for posting our Cosmos-Transfer1! The code is released at github.com/nvidia-cosmos/cosmos-transfer1 and model is on hugging face!
AK
@_akhaliq
Nvidia just released Cosmos-Transfer1 on Hugging Face
Conditional World Generation with Adaptive Multimodal ControlThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
NVIDIA
@nvidia
For 16 years running, NVIDIA technologies have powered every Academy Award-nominated film for Best #vfx. This year, NVIDIA researchers are honored for their groundbreaking work in simulation, denoising, and rendering. #gtc25 blogs.nvidia.com/blog/academy-award-scientific-technical-…
-
Sanja Fidler reposted
AK
@_akhaliq
Nvidia just dropped GEN3C
3D-Informed World-Consistent Video Generation with Precise Camera ControlThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Xuanchi Ren
@xuanchi13
🚀Excited to introduce GEN3C #cvpr2025, a generative video model with an explicit 3D cache for precise camera control.
🎥It applies to multiple use cases, including single-view and sparse-view NVS🖼️ and challenging settings like monocular dynamic NVS and driving simulation🚗.
Project page: research.nvidia.com/labs/toronto-ai/GEN3C/ -
Sanja Fidler reposted
Jay Wu
@jayzhangjiewu
Excited to share our #cvpr2025 paper: Difix3D+
Difix3D+ reimagines 3D reconstruction with single-step diffusion, distilling 2D generative priors for realistic novel view synthesis from large viewpoint shifts.
📄Paper: arxiv.org/abs/2503.01774
🌐Website: research.nvidia.com/labs/toronto-ai/difix3d -
Sanja Fidler reposted
Zian Wang
@zianwang97
🚀 Introducing DiffusionRenderer, a neural rendering engine powered by video diffusion models.
🎥 Estimates high-quality geometry and materials from videos, synthesizes photorealistic light transport, enables relighting and material editing with realistic shadows and reflectionsThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Huan Ling
@HuanLing6
I am so proud to be part of this big project. Checkout Cosmos models and see how powerful it is to finetune to physical AI downstream applications. More examples at research.nvidia.com/labs/dir/cosmos1/
Ming-Yu Liu
@liu_mingyu
github.com/NVIDIA/Cosmos
Cosmos is a developer-first platform designed to help physical AI builders accelerate their development. It has pre-trained world foundation models (diffusion & autoregressive) in different sizes and video tokenizers. They are open models with permissive licenses. Try it out, and let us know how we can better help. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Just announced at #ces2025: 👀 Build custom world models for physical #ai systems at scale.
Get started with #nvidiacosmos ➡️ nvda.ws/3PrO6y6
These new open world foundation models are available on NGC and @HuggingFace. -
Sanja Fidler reposted
Ming-Yu Liu
@liu_mingyu
github.com/NVIDIA/Cosmos
Cosmos is a developer-first platform designed to help physical AI builders accelerate their development. It has pre-trained world foundation models (diffusion & autoregressive) in different sizes and video tokenizers. They are open models with permissive licenses. Try it out, and let us know how we can better help. -
Sanja Fidler reposted
Qi Wu
@wilson_over
Say goodbye to perfect pinhole assumptions
Excited to introduce 3DGUT—a Gaussian Splatting formulation that unlocks support for distorted cameras, including time dependent effects like rolling shutter, while maintaining the benefits of rasterization, rendering at >250 FPS. 🧵 -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
👀 New from #nvidiaresearch: LLaMA-Mesh, a groundbreaking approach that enables LLMs to generate 3D meshes.✨🏗️
✅ Leverages spatial knowledge embedded in LLMs
✅ Enables conversational 3D generation and mesh understanding
✅ Represents 3D mesh data as plain text for seamless LLM integration
✅ Achieves mesh generation quality on par with specialized models
This innovation unifies 3D and text modalities, opening up new possibilities for creative AI applications. Imagine describing a #3d object and having an AI generate it instantly.
🙌 Demo: huggingface.co/spaces/Zhengyi/LLaMA-Mesh
📗 Project page: research.nvidia.com/labs/toronto-ai/LLaMA-Mesh/ -
Sanja Fidler reposted
Jonathan Lorraine
@jonLorraine9
🦙New #nvidia paper: LLaMA-Mesh 🦙
We enable LLMs to generate 3D meshes by representing them as plain text and fine-tuning, unifying 3D and text modalities in a single model.
🔎 Project Page: research.nvidia.com/labs/toronto-ai/LLaMA-Mesh/
🕹️ Interactive Demo: huggingface.co/spaces/Zhengyi/LLaMA-Mesh
📖 Full Paper: arxiv.org/abs/2411.09595
👨💻Code: github.com/nv-tlabs/LLaMa-Mesh
💾 Model Checkpoint: huggingface.co/Zhengyi/LLaMA-Mesh
🧩 Blender Addon: github.com/huggingface/meshgen (courtesy of @dylan_ebert_)
🎥 5-min Overview Video: youtu.be/eZNazN-1lPo (courtesy of @aipapersacademy) -
Sanja Fidler reposted
Xuanchi Ren
@xuanchi13
📢🚗✨ Excited to announce InfiniCube, our scalable generative model for dynamic 3D driving scene generation with high fidelity and controllability!
InfiniCube generates very large-scale (300m×400m ~ 100,000m^2), dynamic 3D driving scenes given HD maps, 3D bounding boxes, and text prompts as controls.
Website: research.nvidia.com/labs/toronto-ai/infinicube/
Props to @YifanLu17525599 for leading this project!!!!
[1/N] -
Sanja Fidler reposted
Jiawei Ren
@jiawei6_ren
Excited to introduce BTimer: a real-time solution for reconstructing dynamic scenes from monocular videos! 🚀
Generate per-frame (bullet-time) 3D GS scenes in 150ms on a single GPU, achieving quality that is on par with the optimization methods!
Project: research.nvidia.com/labs/toronto-ai/bullet-timer/ -
Sanja Fidler reposted
AK
@_akhaliq
Nvidia presents LLaMA-Mesh
Unifying 3D Mesh Generation with Language Models -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
Our 3D Gaussian Ray Tracing work was featured on @twominutepapers, check out the cool video:
www.youtube.com/watch?v=UwL-4LOhxx8&ab_channel=TwoMinuteP…
Ashkan Mirzaei
@ashmrz10
⚡ Ray Tracing + 3D Gaussians = New Possibilities!
Gaussian splatting is limited by rasterization—our #siggraphasia2024 paper shows how to ray trace instead, enabling reflections, shadows, fisheye cameras, and more. The most important (and hardest!) part is making it fast. (1/N) -
Sanja Fidler reposted
Sara Oblak
@sara_oblak
Reconstructing dynamic scenes from video inputs is challenging due to the sparse nature of inputs, both in time and space. In our new preprint, we address this issue by introducing the ReMatching framework - a novel approach for designing and integrating deformation priors into dynamic reconstruction models.
Joint work with @paschalidoud_1 @FidlerSanja @matanatzmon (1/N) -
Sanja Fidler reposted
Nick Sharp
@nmwsharp
We found a way to generate manifold, polygonal meshes from feature vectors at points -- even if the vectors are random, you are still guaranteed to get a manifold mesh!
How? Halfedge meshes, permutations, spacetimes, and more! Check out the 🧵. This project was a blast!
Tianchang Shen
@TianchangS
Generating nice meshes in AI pipelines is hard. Our #siggraphasia2024 paper proposes a new representation which guarantees manifold connectivity, and even supports polygonal meshes -- a big step for downstream editing and simulation. (1/N)
SpaceMesh: research.nvidia.com/labs/toronto-ai/space-mesh/
Joint work with @mli0603 @MarcTLaw @matanatzmon @FidlerSanja @james_r_lucas @JunGao33210520 @nmwsharp (1/n) -
Sanja Fidler reposted
Tianchang Shen
@TianchangS
Generating nice meshes in AI pipelines is hard. Our #siggraphasia2024 paper proposes a new representation which guarantees manifold connectivity, and even supports polygonal meshes -- a big step for downstream editing and simulation. (1/N)
SpaceMesh: research.nvidia.com/labs/toronto-ai/space-mesh/
Joint work with @mli0603 @MarcTLaw @matanatzmon @FidlerSanja @james_r_lucas @JunGao33210520 @nmwsharp (1/n) -
Sanja Fidler reposted
Jonathan Lorraine
@jonLorraine9
Our group is looking for PhD Research Interns for 2025 🔍
The NVIDIA Toronto AI Labs' interests lie at the intersection of machine learning, computer vision, and physics. I'm also interested in foundational ML work, building tools to power AI models of the future.
View more from our lab: research.nvidia.com/labs/toronto-ai/
Apply here: nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAExternalCareerSi… -
Sanja Fidler reposted
David Levin (not the hockey player)
@diwlevin
The NVIDIA HiFi Physics team is looking for interns with interest/skills in
1⃣ Physics simulation (solids/fluids/differentiable etc)
2⃣ Digital Humans
3⃣ Neural Rendering/Geometry, Video Models
4⃣ Monte Carlo Methods
5⃣ Digital Fabrication
Apply: nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAExternalCareerSi… -
Sanja Fidler reposted
Ashkan Mirzaei
@ashmrz10
⚡ Ray Tracing + 3D Gaussians = New Possibilities!
Gaussian splatting is limited by rasterization—our #siggraphasia2024 paper shows how to ray trace instead, enabling reflections, shadows, fisheye cameras, and more. The most important (and hardest!) part is making it fast. (1/N) -
Sanja Fidler reposted
AK
@_akhaliq
Photorealistic Object Insertion with Diffusion-Guided Inverse Rendering
discuss: huggingface.co/papers/2408.09702
The correct insertion of virtual objects in images of real-world scenes requires a deep understanding of the scene's lighting, geometry and materials, as well as the image formation process. While recent large-scale diffusion models have shown strong generative and inpainting capabilities, we find that current models do not sufficiently "understand" the scene shown in a single picture to generate consistent lighting effects (shadows, bright reflections, etc.) while preserving the identity and details of the composited object. We propose using a personalized large diffusion model as guidance to a physically based inverse rendering process. Our method recovers scene lighting and tone-mapping parameters, allowing the photorealistic composition of arbitrary virtual objects in single frames or videos of indoor or outdoor scenes. Our physically based pipeline further enables automatic materials and tone-mapping refinement. -
Sanja Fidler reposted
Ruofan Liang
@RfLiang
🌟 Seamlessly insert 3D objects into any scene!
🚀 Introducing our #eccv2024 work “DiPIR”, a method to recover lighting from a single image, enabling photorealistic virtual object compositing into indoor and outdoor scenes.
Project page: research.nvidia.com/labs/toronto-ai/DiPIR/ -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
📣 Attention Developers!
Attend the #openusd Developer Meetup at #siggraph2024 to build lasting relationships with other industry leaders, share insights, and inspire each other through shared experiences.
nvda.ws/3zHpKvD -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
Looking for hands-on #openusd training?
Don't miss out on the chance to experience our in-person training labs at #siggraph2024. 💻
nvda.ws/3zM16d8 -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
OpenUSD Day is coming to #siggraph2024 👀
Register today & stay tuned as we reveal sessions covering how developers & industry leaders are adopting #openusd to build #ai-enabled 3D pipelines on Tuesday, July 30 👉 nvda.ws/3WbBPlL -
Sanja Fidler reposted
Francis Williams
@frncswllms
📣📣📣 We've officially released the code for XCube, a method for scalable 3D generation, here: github.com/nv-tlabs/XCube
We're also presenting XCube at #cvpr2024 tomorrow at 10:00AM PDT! Come check out our poster session and ask us questions!
Francis Williams
@frncswllms
📢📢📢 Check out XCube, a generative model for very-high resolution 3D shapes and scenes!
Project Page: research.nvidia.com/labs/toronto-ai/xcube/
Project Video: youtu.be/GX0lzwy8nUI
Props to @xuanchi13 for leading this project!!!!
[1/N] -
Sanja Fidler reposted
Jiawei Ren
@jiawei6_ren
We present #l4gm, the first 4D Large Reconstruction Model that produces animated objects from a single-view video input -- in a single feed-forward pass that takes only **seconds**! research.nvidia.com/labs/toronto-ai/l4gm/ 1/
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Today marks a huge milestone for @Waabi – we’ve raised $200M USD in an oversubscribed Series B round, led by @Uber & @khoslaventures. This significant investment will accelerate our progress toward launching fully driverless trucks in 2025!
waabi.ai/waabi-series-b-announcement/ -
Sanja Fidler reposted
Boyi Li
@Boyiliee
🤖 Our "Vision and Language for Autonomous Driving and Robotics" full-day workshop @CVPR will take place next Tuesday. Please check the details here: vision-language-adr.github.io
See you in Seattle!
Boyi Li
@Boyiliee
We are happy to announce the first Vision and Language for Autonomous Driving and Robotics (VLADR) workshop at @CVPR 2024!
Call for contributions and more details 👇🏻
vision-language-adr.github.io
See you in Seattle! 😃 -
Sanja Fidler reposted
Pavlo Molchanov
@PavloMolchanov
🚨Update on RADIO - a single vision foundation model to rule them all! You get CLIP, DINO, SAM features in a single pass! Like a Swiss army knife for CV.
Code: github.com/NVlabs/RADIO
Paper: arxiv.org/abs/2312.06709
- Accepted at CVPR2024
- Arxiv update
- V2.1 model with better metrics: kNN 86.06, ADE20k 51.34 (linear)
- Out of the box 3D awarness (depth, normals, multi-view)
- PCA features are less noisy
- BF16 weights for easier VLLM integration -
Sanja Fidler reposted
AK
@_akhaliq
NeRF-XL
Scaling NeRFs with Multiple GPUs
We present NeRF-XL, a principled method for distributing Neural Radiance Fields (NeRFs) across multiple GPUs, thus enabling the training and rendering of NeRFs with an arbitrarily large capacity. We begin by revisiting -
Sanja Fidler reposted
Ruilong Li
@ruilong_li
Excited to release 🚀NeRF-XL🚀, a principled method to distribute your NeRF's parameters to multiple GPUs!
research.nvidia.com/labs/toronto-ai/nerfxl/
It is heuristics-free and mathematically equiv. with training your NeRF on a single (large) GPU.
Let's scale it up!🌎This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
📢📢Most diffusion (and flow matching) models use handcrafted schedules for their denoising steps during sampling. We show how to optimize them in a principled manner for high-quality generation!
@amsabour added quickstart guide & collab to get you started quickly (links below)!
Amirmojtaba Sabour
@amsabour
📢📢 Align Your Steps: Optimizing Sampling Schedules in Diffusion Models
research.nvidia.com/labs/toronto-ai/AlignYourSteps/
TL;DR: We introduce a method for obtaining improved sampling schedules for diffusion models, resulting in better samples at the same computation cost.
(1/5) -
Sanja Fidler reposted
AK
@_akhaliq
Nvidia presents Align Your Steps
Optimizing Sampling Schedules in Diffusion Models
Diffusion models (DMs) have established themselves as the state-of-the-art generative modeling approach in the visual domain and beyond. A crucial drawback of DMs is their slow sampling speed, -
Sanja Fidler reposted
Radiance Fields
@RadianceFields
RefFusion from NVIDIA uses reference images to effectively inpaint, insert objects, and outpaint with Gaussian Splatting.
🔗radiancefields.com/reffusion-inpainting-with-3dgs -
Sanja Fidler reposted
Amirmojtaba Sabour
@amsabour
📢📢 Align Your Steps: Optimizing Sampling Schedules in Diffusion Models
research.nvidia.com/labs/toronto-ai/AlignYourSteps/
TL;DR: We introduce a method for obtaining improved sampling schedules for diffusion models, resulting in better samples at the same computation cost.
(1/5) -
Sanja Fidler reposted
Hongwei Yi
@HongweiYi2
We present TeSMo (research.nvidia.com/labs/toronto-ai/tesmo/
), a text-controlled scene-aware motion generation method based on denoising diffusion models. It’s an exciting collaboration with @JustusThies, @Michael_J_Black, @xbpeng4, @davrempe. (1/7) -
Sanja Fidler reposted
Davis Rempe
@davrempe
We build on the recent success of (scene-unaware) text-to-motion diffusion models by showing how to fine-tune them to generate realistic scene interactions. This work was led by @HongweiYi2 during his internship @NVIDIAAI!
Hongwei Yi
@HongweiYi2
We present TeSMo (research.nvidia.com/labs/toronto-ai/tesmo/
), a text-controlled scene-aware motion generation method based on denoising diffusion models. It’s an exciting collaboration with @JustusThies, @Michael_J_Black, @xbpeng4, @davrempe. (1/7) -
Sanja Fidler reposted
Phillip Isola
@phillip_isola
Our computer vision textbook is released!
Foundations of Computer Vision
with Antonio Torralba and Bill Freeman
mitpress.mit.edu/9780262048972/foundations-of-computer-vi…
It’s been in the works for >10 years. Covers everything from linear filters and camera optics to diffusion models and radiance fields.
1/4 -
Sanja Fidler reposted
Ashkan Mirzaei
@ashmrz10
📢We introduce “RefFusion”, a novel inpainting method for scenes reconstructed using 3D Gaussian Splatting.
🔗reffusion.github.io
TLDR: we personalize an image diffusion model to a given reference image and distill its knowledge to 3D through score distillation sampling. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
🏗️ 📐 Tired of marching cubes producing skinny triangles and losing sharp geometric details?
Check out the latest NVIDIA Kaolin tutorial to learn how to add Flexicubes for mesh optimization from #nvidiaresearch.
👀 www.youtube.com/watch?v=2gqdsUHBhEIThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
In our full-day #cvpr2024 workshop on data-driven simulation for Autonomous Driving, we are bringing together experts from perception, behavior and traffic modeling, and safety to discuss how simulation can benefit real-world AV production pipelines.
agents4ad.github.io -
Sanja Fidler reposted
AK
@_akhaliq
Nvidia presents LATTE3D
Large-scale Amortized Text-To-Enhanced 3D Synthesis
Recent text-to-3D generation approaches produce impressive
3D results but require time-consuming optimization that can take up to
an hour per prompt [21, 39]. Amortized methods like ATT3D [26] -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
High-quality text-to-3D in 400ms! Great work by @jonLorraine9 and the team! @NVIDIAAI
Jonathan Lorraine
@jonLorraine9
New #nvidia #gtc24 paper 🎊
We generate high-quality 3D assets in only 400ms from text by combining (a) amortized optimization for speed, (b) surface rendering for quality, and (c) 3D data for robustness.
☕ LATTE3D project details: research.nvidia.com/labs/toronto-ai/LATTE3D/
🧵with many fun gifs -
Sanja Fidler reposted
Jonathan Lorraine
@jonLorraine9
New #nvidia #gtc24 paper 🎊
We generate high-quality 3D assets in only 400ms from text by combining (a) amortized optimization for speed, (b) surface rendering for quality, and (c) 3D data for robustness.
☕ LATTE3D project details: research.nvidia.com/labs/toronto-ai/LATTE3D/
🧵with many fun gifs -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
☕ The new LATTE3D model from #nvidiaresearch turns text prompts into high-quality 3D representations of objects and animals within a second on a single RTX GPU. #gtc24
👀 nvda.ws/3TzYN3s -
Sanja Fidler reposted
NVIDIA GTC
@NVIDIAGTC
Only 11 days left until #gtc24. Don't miss NVIDIA CEO Jensen Huang's highly anticipated keynote live onstage March 18 at 1 p.m. PDT.
💫Discover the latest technology driving the #generativeai revolution and beyond. nvda.ws/4c6XoJV -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
On #internationalwomensday, I am feeling immense pride in the culture we've built at @Waabi_ai . Here's to all the incredible women of Waabi and the world who inspire me daily.
Waabi
@Waabi_ai
Happy International Women's Day! At Waabi, we are proud to foster a culture where women can thrive, today and every day. Hear from some of the incredible women at Waabi in this video:
-
Sanja Fidler reposted
-
Sanja Fidler reposted
Boyi Li
@Boyiliee
We are happy to announce the first Vision and Language for Autonomous Driving and Robotics (VLADR) workshop at @CVPR 2024!
Call for contributions and more details 👇🏻
vision-language-adr.github.io
See you in Seattle! 😃 -
Sanja Fidler reposted
Kosta Derpanis
@CSProfKGD
It’s @nvidia week here in #toronto. Next up, @karsten_kreis giving a @UofTCompSci lecture on generative modeling 🤗
-
Sanja Fidler reposted
Igor Gilitschenski
@igilitschenski
Together with @DaveLindell, we are excited to host @NVIDIAAI‘s @ZGojcic for a talk at the Toronto Vision Seminar!
-
Sanja Fidler reposted
AK
@_akhaliq
Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation
paper page: huggingface.co/papers/2401.08559
Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text, with methods that can generate character animations from short prompts and specified durations. However, using a single text prompt as input lacks the fine-grained control needed by animators, such as composing multiple actions and defining precise durations for parts of the motion. To address this, we introduce the new problem of timeline control for text-driven motion synthesis, which provides an intuitive, yet fine-grained, input interface for users. Instead of a single prompt, users can specify a multi-track timeline of multiple prompts organized in temporal intervals that may overlap. This enables specifying the exact timings of each action and composing multiple actions in sequence or at overlapping intervals. To generate composite animations from a multi-track timeline, we propose a new test-time denoising method. This method can be integrated with any pre-trained motion diffusion model to synthesize realistic motions that accurately reflect the timeline. At every step of denoising, our method processes each timeline interval (text prompt) individually, subsequently aggregating the predictions with consideration for the specific body parts engaged in each action. Experimental comparisons and ablations validate that our method produces realistic motions that respect the semantics and timing of given text prompts. -
Sanja Fidler reposted
Davis Rempe
@davrempe
Check out our recent work led by @MathisPetrovich that generates human motions from a timeline of text prompts, similar to a typical video editor.
The method operates entirely at test time, so it works with off-the-shelf motion diffusion models!
Project: mathis.petrovich.fr/stmc/
AK
@_akhaliq
Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation
paper page: huggingface.co/papers/2401.08559
Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text, with methods that can generate character animations from short prompts and specified durations. However, using a single text prompt as input lacks the fine-grained control needed by animators, such as composing multiple actions and defining precise durations for parts of the motion. To address this, we introduce the new problem of timeline control for text-driven motion synthesis, which provides an intuitive, yet fine-grained, input interface for users. Instead of a single prompt, users can specify a multi-track timeline of multiple prompts organized in temporal intervals that may overlap. This enables specifying the exact timings of each action and composing multiple actions in sequence or at overlapping intervals. To generate composite animations from a multi-track timeline, we propose a new test-time denoising method. This method can be integrated with any pre-trained motion diffusion model to synthesize realistic motions that accurately reflect the timeline. At every step of denoising, our method processes each timeline interval (text prompt) individually, subsequently aggregating the predictions with consideration for the specific body parts engaged in each action. Experimental comparisons and ablations validate that our method produces realistic motions that respect the semantics and timing of given text prompts. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Exciting news for #pytorch enthusiasts: Our NVIDIA Kaolin library introduces SurfaceMesh class to simplify managing mesh attributes including consistency, auto-compute normals, & streamline indexing.
👀 See the video tutorial from #nvidiaresearch
➡️ nvda.ws/4aRCi1D -
Sanja Fidler reposted
Masha Shugrina
@_shumash
Check out this convenience class for handling surface mesh tensor attributes. We hope it's useful and saves you at least some of the headache! #pytorch #ai #deeplearning #3d
NVIDIA AI Developer
@NVIDIAAIDev
Exciting news for #pytorch enthusiasts: Our NVIDIA Kaolin library introduces SurfaceMesh class to simplify managing mesh attributes including consistency, auto-compute normals, & streamline indexing.
👀 See the video tutorial from #nvidiaresearch
➡️ nvda.ws/4aRCi1D -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I am so proud to be appointed to the Order of Ontario! It is an honor to be recognized alongside such outstanding Ontarians who have made exceptional contributions toward a better province, country, and world. Congratulations to my fellow 24 appointees!
news.ontario.ca/en/release/1004014/province-honours-25-ne… -
Sanja Fidler reposted
Jonah Philion
@PhilionJonah
Very cool to see Trajeglish share the stage for a minute with @StabilityAI SDXL Turbo in last week's @twominutepapers video!
video: youtu.be/Iol2rb65aSk?t=339
Trajeglish: research.nvidia.com/labs/toronto-ai/trajeglish/
1/3 -
Sanja Fidler reposted
AK
@_akhaliq
Nvidia announces Compact NGP
Compact Neural Graphics Primitives with Learned Hash Probing
paper page: huggingface.co/papers/2312.17241
Up to 4x smaller than Instant NGP at 1.2-2.6x training cost and no inference speed penalty -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
👀 Discover the top 10 #nvidiaresearch projects of the year.
✨ From Neuralangelo's high-fidelity neural surface reconstruction to Magic3D's text-to-3D content creation, these projects push the boundaries of innovation in #ai.
nvda.ws/3RlypJr -
Sanja Fidler reposted
Masha Shugrina
@_shumash
So proud of my team for presenting the first interactive #texture #painting with #ai at #siggraphasia2023 Real-Time Live. Well done Anita Hu and team!!
We want the artist to stay in control 🎨🖌️🤗
developer.nvidia.com/blog/nvidia-research-shows-interacti… -
Sanja Fidler reposted
U of T Department of Computer Science
@UofTCompSci
Computer graphics research from @UofTCompSci is well represented at #siggraphasia2023, with 14 accepted technical papers, including a Best Paper win! uoft.me/9-C
-
Sanja Fidler
@FidlerSanja
Very proud of the amazing team and collaborators at @NVIDIAAI to win the Best Paper Award at #siggraphasia2023!
Zian Wang
@zianwang97
Proud to share that Adaptive Shells has been selected for a SIGGRAPH Asia Best Paper Award! #siggraphasia2023 #nvidia
Zian Wang
@zianwang97
🚀 Introducing our #siggraphasia work “Adaptive Shells”, a novel #nerf formulation that yields high visual fidelity and greatly accelerates rendering.
TLDR: Auto-derived bounding shells result in up to 10x faster inference than InstantNGP!
[1/n]This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Jonah Philion
@PhilionJonah
Trajeglish tops the Waymo Sim Agents Benchmark, outperforming several well-established motion prediction models (Wayformer, MutliPath++, MTR). Our model sets a new SOTA along the leaderboard realism metric by 3.3% and the interaction metric by 9.9%
5/6 -
Sanja Fidler reposted
Jiahui Huang
@huangjh_hjh
Glad to present XCube altogether with @xuanchi13 Xiaohui, Ken, @FidlerSanja, and @frncswllms!
XCube is a 3D diffusion model that could generate high-resolution geometries at a scene scale (100m x 100m).
🔥Sampling time is ultra-fast thanks to our sparse voxel representation.
NVIDIA AI Developer
@NVIDIAAIDev
📣 #nvidiaresearch paper, XCube, generates highly-detailed #3d scenes with resolutions of up to 1024³ in under 30 seconds.
👀 Capable of leveraging voxel attributes such as normals, semantics, & TSDF, this model is pushing the boundary of 3D generation.
nvda.ws/3TeXk3KThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
📣 #nvidiaresearch paper, XCube, generates highly-detailed #3d scenes with resolutions of up to 1024³ in under 30 seconds.
👀 Capable of leveraging voxel attributes such as normals, semantics, & TSDF, this model is pushing the boundary of 3D generation.
nvda.ws/3TeXk3KThis media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Jonah Philion
@PhilionJonah
"Trajeglish: Learning the Language of Driving Scenarios" w/ @xbpeng4 @FidlerSanja
Discrete sequence modeling for controlling interactive agents in self-driving simulation
@nvidia @VectorInst @UofTCompSci @SFU
research.nvidia.com/labs/toronto-ai/trajeglish/
arxiv.org/abs/2312.04535
1/6 -
Sanja Fidler reposted
Francis Williams
@frncswllms
📢📢📢 Check out XCube, a generative model for very-high resolution 3D shapes and scenes!
Project Page: research.nvidia.com/labs/toronto-ai/xcube/
Project Video: youtu.be/GX0lzwy8nUI
Props to @xuanchi13 for leading this project!!!!
[1/N] -
Sanja Fidler reposted
AK
@_akhaliq
Adaptive Shells for Efficient Neural Radiance Field Rendering
paper page: huggingface.co/papers/2311.10091
Neural radiance fields achieve unprecedented quality for novel view synthesis, but their volumetric formulation remains expensive, requiring a huge number of samples to render high-resolution images. Volumetric encodings are essential to represent fuzzy geometry such as foliage and hair, and they are well-suited for stochastic optimization. Yet, many scenes ultimately consist largely of solid surfaces which can be accurately rendered by a single sample per pixel. Based on this insight, we propose a neural radiance formulation that smoothly transitions between volumetric- and surface-based rendering, greatly accelerating rendering speed and even improving visual fidelity. Our method constructs an explicit mesh envelope which spatially bounds a neural volumetric representation. In solid regions, the envelope nearly converges to a surface and can often be rendered with a single sample. To this end, we generalize the NeuS formulation with a learned spatially-varying kernel size which encodes the spread of the density, fitting a wide kernel to volume-like regions and a tight kernel to surface-like regions. We then extract an explicit mesh of a narrow band around the surface, with width determined by the kernel size, and fine-tune the radiance field within this band. At inference time, we cast rays against the mesh and evaluate the radiance field only within the enclosed region, greatly reducing the number of samples required. Experiments show that our approach enables efficient rendering at very high fidelity. We also demonstrate that the extracted envelope enables downstream applications such as animation and simulation.This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Zian Wang
@zianwang97
🚀 Introducing our #siggraphasia work “Adaptive Shells”, a novel #nerf formulation that yields high visual fidelity and greatly accelerates rendering.
TLDR: Auto-derived bounding shells result in up to 10x faster inference than InstantNGP!
[1/n]This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Chenfeng_X
@Chenfeng_X
Can StableDiff go beyond image generation? 🤔️
📢Our latest research, 3DiffTection, takes StableDiffusion to new heights by enabling 3D object detection from a single image. @orlitany @FidlerSanja @HuanLing6 See our project page research.nvidia.com/labs/toronto-ai/3difftection/ -
Sanja Fidler reposted
AK
@_akhaliq
3DiffTection: 3D Object Detection with Geometry-Aware Diffusion Features
paper page: huggingface.co/papers/2311.04391
present 3DiffTection, a state-of-the-art method for 3D object detection from single images, leveraging features from a 3D-aware diffusion model. Annotating large-scale image data for 3D detection is resource-intensive and time-consuming. Recently, pretrained large image diffusion models have become prominent as effective feature extractors for 2D perception tasks. However, these features are initially trained on paired text and image data, which are not optimized for 3D tasks, and often exhibit a domain gap when applied to the target data. Our approach bridges these gaps through two specialized tuning strategies: geometric and semantic. For geometric tuning, we fine-tune a diffusion model to perform novel view synthesis conditioned on a single image, by introducing a novel epipolar warp operator. This task meets two essential criteria: the necessity for 3D awareness and reliance solely on posed image data, which are readily available (e.g., from videos) and does not require manual annotation. For semantic refinement, we further train the model on target data with detection supervision. Both tuning phases employ ControlNet to preserve the integrity of the original feature capabilities. In the final step, we harness these enhanced capabilities to conduct a test-time prediction ensemble across multiple virtual viewpoints. Through our methodology, we obtain 3D-aware features that are tailored for 3D detection and excel in identifying cross-view point correspondences. Consequently, our model emerges as a powerful 3D detector, substantially surpassing previous benchmarks, e.g., Cube-RCNN, a precedent in single-view 3D detection by 9.43\% in AP3D on the Omni3D-ARkitscene dataset. Furthermore, 3DiffTection showcases robust data efficiency and generalization to cross-domain data. -
Sanja Fidler reposted
Jonathan Lorraine
@jonLorraine9
New #nvidia paper: Real-time text-to-3D generation #iccv2023
3D generation from text requires expensive per-prompt optimization. We train 1 model on many prompts for real-time generalization to unseen prompts, interpolations and more!
ATT3D details:
research.nvidia.com/labs/toronto-ai/ATT3D/ -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
So excited to share @Waabi_ai strategic partnership with @UberFreight to accelerate the safe deployment of #ai-powered autonomous trucks at scale. Huge step toward the future of safer roads and more efficient supply chains. waabi.ai/waabi-uber-freight/
-
Sanja Fidler reposted
Tianchang Shen
@TianchangS
📢We released the code for FlexiCubes! Check out our tutorials and examples to discover how to incorporate FlexiCubes in your mesh optimization pipelines🔨
github.com/nv-tlabs/FlexiCubes
Tianchang Shen
@TianchangS
Excited to share our #siggraph2023 project for better mesh extraction in learning & reconstruction pipelines.
Project page: rb.gy/8nypr
Joint work w/ @yoslber Jon Hasselgren @kangxue_yin @zianwang97 @ChenWenzheng @ZGojcic @FidlerSanja @nmwsharp @JunGao33210520
(1/n) -
Sanja Fidler reposted
NVIDIA DRIVE
@NVIDIADRIVE
3D occupancy prediction for #autonomousdriving provides rich information about the world to an #av planning stack. Watch the latest episode of #drivelabs to learn more. nvda.ws/3QTCPZf
-
Sanja Fidler reposted
AK
@_akhaliq
Flexible Isosurface Extraction for Gradient-Based Mesh Optimization
paper page: huggingface.co/papers/2308.05371
This work considers gradient-based mesh optimization, where we iteratively optimize for a 3D surface mesh by representing it as the isosurface of a scalar field, an increasingly common paradigm in applications including photogrammetry, generative modeling, and inverse physics. Existing implementations adapt classic isosurface extraction algorithms like Marching Cubes or Dual Contouring; these techniques were designed to extract meshes from fixed, known fields, and in the optimization setting they lack the degrees of freedom to represent high-quality feature-preserving meshes, or suffer from numerical instabilities. We introduce FlexiCubes, an isosurface representation specifically designed for optimizing an unknown mesh with respect to geometric, visual, or even physical objectives. Our main insight is to introduce additional carefully-chosen parameters into the representation, which allow local flexible adjustments to the extracted mesh geometry and connectivity. These parameters are updated along with the underlying scalar field via automatic differentiation when optimizing for a downstream task. We base our extraction scheme on Dual Marching Cubes for improved topological properties, and present extensions to optionally generate tetrahedral and hierarchically-adaptive meshes. Extensive experiments validate FlexiCubes on both synthetic benchmarks and real-world applications, showing that it offers significant improvements in mesh quality and geometric fidelity. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
📣 Just released: #nvidiakaolin v0.14.0
Accelerate 3D deep learning research, debug your custom renderer interactively in a Jupyter notebook, and manage batched mesh attributes with a SurfaceMesh container, within the Kaolin #pytorch library.
Github: nvda.ws/3QeuDCr -
Sanja Fidler reposted
Huan Ling
@HuanLing6
Come check out our #cvpr2023 paper VideoLDM(research.nvidia.com/labs/toronto-ai/VideoLDM/) at the PM poster session today (4:30PM) at poster #187!!!
w/ @andi_blatt,@robrombach,@timudk, @seungkim0123,@FidlerSanja,@karsten_kreis
We make an animal band using our model!
@NVIDIAAI
Karsten Kreis
@karsten_kreis
📢📢 Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
research.nvidia.com/labs/toronto-ai/VideoLDM/
We turn image LDMs into high-res. video generators! #cvpr2023
w/ the amazing @andi_blatt*, @robrombach*, @HuanLing6*, @timudk*, @seungkim0123, @FidlerSanja
(1/n) -
Sanja Fidler reposted
Francis Williams
@frncswllms
#cvpr2023 For those who haven't tried yet: PCU is a fast and easy-to-use Python lib for processing geometric data! It's great for cleaning up 3D datasets for Deep Learning pipelines!
Docs: www.fwilliams.info/point-cloud-utils/
GitHub: github.com/fwilliams/point-cloud-utils
Some things it can do 🧵: -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
🎉 Congrats to our #nvidiaresearch team for winning 🏆 1st Place and the 🏆 Innovation Award for the #cvpr2023 #3d Occupancy Challenge.
Their 3D object detection method substantially improved performance in comparison to prior approaches.
Read the paper: nvda.ws/3CoXTym -
Sanja Fidler reposted
Or Litany
@orlitany
Thrilled to announce our publication - making #ai surface reconstruction industry-ready! NKF proved learnable kernels' power for varied densities & novel categories. NKSR tackles the hurdles of scale & noise robustness - now we're truly real-world ready. #surfacereconstruction🚀
Francis Williams
@frncswllms
Want to reconstruct a mesh from 300m points in 2mins on a small GPU cluster? Check out NKSR, our CVPR highlight paper on 3D reconstruction. NKSR works on arbitrary noisy point clouds out of the box! How does it work?
research.nvidia.com/labs/toronto-ai/NKSR/
Kudos to lead author @huangjh_hjh!
🧵 -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
For the past two years, we've been building the foundation of a new, transformative approach to self-driving. Research and innovation are central to this approach, and today I'm excited to unveil that research to the world.
waabi.ai/why-purposeful-research-and-innovation-are-the-k… -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Dive into the #code to leverage #neuralnetworks and kernel-based methods to enhance real-world objects' #3d reconstruction process with #nvidiaresearch Neural Kernel Surface Reconstruction.
And maybe win an RTX 3090Ti ...🧵 👇 New weekend plans?
nvda.ws/3NkPYs5This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Francis Williams
@frncswllms
Want to reconstruct a mesh from 300m points in 2mins on a small GPU cluster? Check out NKSR, our CVPR highlight paper on 3D reconstruction. NKSR works on arbitrary noisy point clouds out of the box! How does it work?
research.nvidia.com/labs/toronto-ai/NKSR/
Kudos to lead author @huangjh_hjh!
🧵 -
Sanja Fidler reposted
Jiahui Huang
@huangjh_hjh
📢Check out our newest work Neural Kernel Surface Reconstruction (NKSR) which reconstructs mesh from point clouds in a highly accurate and scalable way!
🍡Project Page: research.nvidia.com/labs/toronto-ai/NKSR/
Francis Williams
@frncswllms
Want to reconstruct a mesh from 300m points in 2mins on a small GPU cluster? Check out NKSR, our CVPR highlight paper on 3D reconstruction. NKSR works on arbitrary noisy point clouds out of the box! How does it work?
research.nvidia.com/labs/toronto-ai/NKSR/
Kudos to lead author @huangjh_hjh!
🧵 -
Sanja Fidler reposted
Mohamed Hassan
@M_E_Hassan
Do you want your characters to interact physically and naturally with the scene? Tired of manual annotation? Want to generalize far beyond the training objects and scenarios? Then look no further to our #siggraph2023 work
xbpeng.github.io/projects/InterPhys/index.html -
Sanja Fidler reposted
Aaron Lefohn
@aaronlefohn
NVIDIA is thrilled to share our SIGGRAPH 2023 papers. It is our best SIGGRAPH yet, with ~20 papers authored in partnership with over a dozen universities.
Generative AI, 3D content creation, neural rendering, neural physics, and more...
blogs.nvidia.com/blog/2023/05/02/graphics-research-advanc… -
Sanja Fidler reposted
AK
@_akhaliq
Learning Physically Simulated Tennis Skills from Broadcast Videos
demonstrate that system produces controllers for physically-simulated tennis players that can hit the incoming ball to target positions accurately using a diverse array of strokes (serves, forehands, and backhands), spins (topspins and slices), and playing styles (one/two-handed backhands, left/right-handed play). Overall, system can synthesize two physically simulated characters playing extended tennis rallies with simulated racket and ball dynamics
paper: research.nvidia.com/labs/toronto-ai/vid2player3d/data/ten…
project page: research.nvidia.com/labs/toronto-ai/vid2player3d/This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Haotian Zhang
@haotianz4
I'm super excited to finally share our work of learning physically simulated tennis skills from broadcast videos (with
@KhrylxYe @FidlerSanja @xbpeng4 @kayvonf
), accepted to #siggraph2023!
Check it out at research.nvidia.com/labs/toronto-ai/vid2player3d/ -
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
📢Toronto Annotation Suite, a web-based annotation platform we've been developing @UofT, just got better with the integration of the Segment Anything Model, enabling even more precise and generalizable AI-assisted annotation processes.
Try it out here!
aidemos.cs.toronto.edu/toras -
Sanja Fidler reposted
Huan Ling
@HuanLing6
We are excited to see people like our videoLDM! As a team of text2video, we have our unique way to celebrate a successful release by generating a video 😆
w/ @andi_blatt,@robrombach,@timudk,@FidlerSanja
,@karsten_kreis
"Storm trooper cheers"
AK
@_akhaliq
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to 1280 x 2048
abs: arxiv.org/abs/2304.08818
project page: research.nvidia.com/labs/toronto-ai/VideoLDM/This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Check out NVIDIA’s newest generative model for complex open-world 3D scenes.
It will generate novel scenes, which you can then edit and style for a wide range of uses, including potential use cases in VR & robotics.
nvda.ws/40l6ZWA -
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
📢Excited to announce "NeuralField-LDM: Scene Generation with Hierarchical Latent Diffusion Models"!
research.nvidia.com/labs/toronto-ai/NFLDM/
Joint work w/ the amazing @brad19brown @kangxue_yin @karsten_kreis @K_S_Schwarz @lidaiqing @robrombach @abtorralba @FidlerSanja
@NVIDIAAI #cvpr2023 -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
📢📢 Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
research.nvidia.com/labs/toronto-ai/VideoLDM/
We turn image LDMs into high-res. video generators! #cvpr2023
w/ the amazing @andi_blatt*, @robrombach*, @HuanLing6*, @timudk*, @seungkim0123, @FidlerSanja
(1/n) -
Sanja Fidler reposted
AK
@_akhaliq
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to 1280 x 2048
abs: arxiv.org/abs/2304.08818
project page: research.nvidia.com/labs/toronto-ai/VideoLDM/This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
AK
@_akhaliq
Neural Fields meet Explicit Geometric Representations for Inverse Rendering of Urban Scenes
abs: arxiv.org/abs/2304.03266
project page: nv-tlabs.github.io/fegr/ -
Sanja Fidler reposted
Jun Gao
@JunGao33210520
Check our CVPR 2023 work!
We combine neural fields with explicit meshes to enable neural reconstruction with geometry, materials, lighting and shadow for urban scenes that support controllable relighting and object insertion!!
Zian Wang
@zianwang97
Excited to share our #cvpr2023 work “Neural Fields meet Explicit Geometric Representations”!
Project: nv-tlabs.github.io/fegr/
Video: youtu.be/1KvHY3tlhhY
Joint work w/ @TianchangS @JunGao33210520 @ShengyHuang @yoslber Jon Hasselgren @ZGojcic @ChenWenzheng @FidlerSanja
NVIDIA AI Developer
@NVIDIAAIDev
1/6 Excited to share our #cvpr2023 work FEGR from @zianwang97
We combine neural fields with explicit meshes to enable neural reconstruction of urban scenes that support controllable relighting and object insertion.
Project: nvda.ws/3zB3dgP
Video: nvda.ws/3zCB8Wb -
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
NeRFs are conquering the scene, but explicit meshes still have a lot to offer. Check out FEGR to learn how we can get the best of both worlds. Amazing work lead by @zianwang97. #cvpr2023
Zian Wang
@zianwang97
Excited to share our #cvpr2023 work “Neural Fields meet Explicit Geometric Representations”!
Project: nv-tlabs.github.io/fegr/
Video: youtu.be/1KvHY3tlhhY
Joint work w/ @TianchangS @JunGao33210520 @ShengyHuang @yoslber Jon Hasselgren @ZGojcic @ChenWenzheng @FidlerSanja
NVIDIA AI Developer
@NVIDIAAIDev
1/6 Excited to share our #cvpr2023 work FEGR from @zianwang97
We combine neural fields with explicit meshes to enable neural reconstruction of urban scenes that support controllable relighting and object insertion.
Project: nvda.ws/3zB3dgP
Video: nvda.ws/3zCB8Wb -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
1/6 Excited to share our #cvpr2023 work FEGR from @zianwang97
We combine neural fields with explicit meshes to enable neural reconstruction of urban scenes that support controllable relighting and object insertion.
Project: nvda.ws/3zB3dgP
Video: nvda.ws/3zCB8Wb -
Sanja Fidler reposted
Zian Wang
@zianwang97
Excited to share our #cvpr2023 work “Neural Fields meet Explicit Geometric Representations”!
Project: nv-tlabs.github.io/fegr/
Video: youtu.be/1KvHY3tlhhY
Joint work w/ @TianchangS @JunGao33210520 @ShengyHuang @yoslber Jon Hasselgren @ZGojcic @ChenWenzheng @FidlerSanja
NVIDIA AI Developer
@NVIDIAAIDev
1/6 Excited to share our #cvpr2023 work FEGR from @zianwang97
We combine neural fields with explicit meshes to enable neural reconstruction of urban scenes that support controllable relighting and object insertion.
Project: nvda.ws/3zB3dgP
Video: nvda.ws/3zCB8Wb -
Sanja Fidler reposted
Davis Rempe
@davrempe
Excited to announce that our work “Trace & Pace” will be presented @CVPR2023.
We combine guided trajectory diffusion with a physics-based humanoid controller to enable pedestrian animation that is controllable by a user.
Project page: nv-tlabs.github.io/trace-pace/
1/5 -
Sanja Fidler reposted
Zhengyi “Zen” Luo
@zhengyiluo
Trace & Pace is live and will be presented at @CVPR 2023!
Project page: nv-tlabs.github.io/trace-pace/
See full thread 👇🏻
Davis Rempe
@davrempe
Excited to announce that our work “Trace & Pace” will be presented @CVPR2023.
We combine guided trajectory diffusion with a physics-based humanoid controller to enable pedestrian animation that is controllable by a user.
Project page: nv-tlabs.github.io/trace-pace/
1/5 -
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
Just launched at #gtc23, NVIDIA AI Foundations, a set of cloud services that provides businesses with a simplified approach to build, customize, and run #generativeai models with their proprietary data for text language, visual content, and biology. nvda.ws/40ntyKx
-
Sanja Fidler reposted
Tsung-Yi Lin
@TsungYiLinCV
#gtc23 revealed the text-to-3D service our team is building. In Jensen’s keynote, we show a city assembled in Omniverse with text-to-3D assets! I am extremely proud to be part of this team for building future tools of 3D content creation!!
www.youtube.com/watch?v=F_dSq1IT6ws&t=103s -
Sanja Fidler reposted
Jun Gao
@JunGao33210520
#gtc2023 revealed our text-to-3D effort! It can generate high-quality 3D shapes (meshes) from texts, with a large improvement upon Magic3D. I’m extremely proud to be part of this fantastic team and build the future of 3D content creation!
www.nvidia.com/en-us/gpu-cloud/picasso/
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
Jensen revealed the @NVIDIAAI text-to-3D (mesh models!) strategy at the GTC Keynote today. I'm super proud to be a part of this epic effort alongside my amazing coworkers 😊
Find out more here: www.nvidia.com/en-us/gpu-cloud/picasso/ -
Sanja Fidler
@FidlerSanja
Very proud of the amazing team & coworkers, and excited to be part of the text to 3D product effort at @NVIDIAAI ! More: nvidia.com/en-us/gpu-cloud/picasso/
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
Jensen revealed the @NVIDIAAI text-to-3D (mesh models!) strategy at the GTC Keynote today. I'm super proud to be a part of this epic effort alongside my amazing coworkers 😊
Find out more here: www.nvidia.com/en-us/gpu-cloud/picasso/ -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
To celebrate #iwd as well two years since the Waabi team began working together, I spent some time reflecting on my life and how my experiences paved the way for me to launch this amazing company alongside so many amazing people. So proud of our team!
www.linkedin.com/pulse/how-my-lifes-experiences-helped-me… -
Sanja Fidler reposted
Yiming Li
@YimingLi9702
Excited to share that our work VoxFormer as a cutting-edge baseline for 3D semantic occupancy prediction from 2D images, is accepted by #cvpr2023! Thank my collaborators and mentors @ZhidingYu @ChrisChoy208 @ChaoweiX @ALVAREZ_JOSEM @FidlerSanja @simbaforrest @AnimaAnandkumar!
-
Sanja Fidler reposted
Fuji Tsang Clément
@Caenorst
Super excited about this release, Kaolin is now installable with a single command line!
Check out our tutorials on lighting:
Diffuse: github.com/NVIDIAGameWorks/kaolin/blob/master/examples/tu…
Specular: github.com/NVIDIAGameWorks/kaolin/blob/master/examples/tu…
Documentation: kaolin.readthedocs.io/en/latest/modules/kaolin.render.lig…
NVIDIA AI Developer
@NVIDIAAIDev
📣 Just released: #nvidiakaolin v0.13.0
With new lighting features enabling #3d DL works like DIBR++ or NeRD, check out our tutorials on diffuse and specular reflectance. Kaolin is now pip installable.
Installation: nvda.ws/3YN40FY
Release notes: nvda.ws/3EbEYZ2 -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
📣 Just released: #nvidiakaolin v0.13.0
With new lighting features enabling #3d DL works like DIBR++ or NeRD, check out our tutorials on diffuse and specular reflectance. Kaolin is now pip installable.
Installation: nvda.ws/3YN40FY
Release notes: nvda.ws/3EbEYZ2 -
Sanja Fidler reposted
Amlan Kar
@amlankar95
📢 We've released V1.1 of the Toronto Annotation Suite!
TORAS is a web-based AI-powered data #annotation platform we have been developing @uoft. We have a number of exciting updates that significantly improve the annotation experience!
Go check it out! aidemos.cs.toronto.edu/toras -
Sanja Fidler reposted
Jordan Juravsky
@jordanjuravsky
Thanks for the tweet @_akhaliq! I'm excited to share PADL, our paper on physics-based text2motion that we presented at SIGGRAPH Asia 2022. More details are on our project page here: nv-tlabs.github.io/PADL/
AK
@_akhaliq
PADL: Language-Directed Physics-Based Character Control
abs: arxiv.org/abs/2301.13868
project page: nv-tlabs.github.io/PADL/ -
Sanja Fidler reposted
AK
@_akhaliq
PADL: Language-Directed Physics-Based Character Control
abs: arxiv.org/abs/2301.13868
project page: nv-tlabs.github.io/PADL/ -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Thrilled to be joining the #volvo Group family.
Waabi
@Waabi_ai
We’re excited to welcome Volvo Group Venture Capital as a strategic investor! Read more here: waabi.ai/welcoming-volvo-group-venture-capital-as-a-strat…
-
Sanja Fidler reposted
Dima Damen
@dimadamen
[RTW- Great Opportunity] Looking 4 PhD student🎓 to join my group @BristolUniEng, on long-form video understanding- in collaboration with @JCollomosse @FabianCabaH.
Open to all &international students - start date from May.
Open until Filled. Apply asap: www.bristol.ac.uk/media-library/sites/engineering/postgra… -
Sanja Fidler reposted
maggie_albrecht
@maggie_albrecht
NVIDIA AI Research Helps Populate Virtual Worlds With 3D Objects | NVIDIA Blog share.nvidia.com/Article/Redirect/a54fb120-f068-45b2-9209…
-
Sanja Fidler reposted
Francis Williams
@frncswllms
📢📢📢📢Want to make an SDF dataset from ShapeNet in 17 lines of code?
Introducing Point Cloud Utils! An easy-to-use Python library for point clouds and meshes. PCU has become my swiss army knife for processing 3D data and can hopefully be yours too! www.fwilliams.info/point-cloud-utils/ -
Sanja Fidler reposted
Exploring Slovenia
@ExploringSlo
Happy New Year 2023!!! Here's a little something waiting for you on the Blog 🤗 Cheers! 🥂www.exploringslovenia.com/blog/2023/01/01/trails-of-2022/
-
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
We have openings for a research engineering internship at NVIDIA to work on open source software (eg github.com/NVIDIAGameWorks/kaolin-wisp) with us. If you want to contrib to cutting edge research for 3D generative modeling, editing and nerfs, send me a DM/email! Undergrads welcome :)
-
Sanja Fidler reposted
Jun Gao
@JunGao33210520
Super excited to see GET3D can work with Objaverse (arxiv.org/abs/2212.08051) !!
Welcome everyone to try GET3D and let me know of any issues in the code!!
github.com/nv-tlabs/GET3D -
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
Our recent SIGGRAPH paper was featured on @twominutepapers!!! Thanks Karoly for the great video, what a time to be alive 🥳
Two Minute Papers
@twominutepapers
New Video - NVIDIA’s New AI: Video Game Graphics, Now 60x Smaller!
youtu.be/aQctoORQwLE
#nvidia #ai -
Sanja Fidler reposted
Two Minute Papers
@twominutepapers
New Video - NVIDIA’s New AI: Video Game Graphics, Now 60x Smaller!
youtu.be/aQctoORQwLE
#nvidia #ai -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
📢📢 Our team at @nvidia (Toronto AI Lab, nv-tlabs.github.io) is looking for strong and motivated PhD research interns!
We are working at the cutting edge in diffusion models, 3D generation and much more.
Email or message me with your CV and website, if you are interested. -
Sanja Fidler reposted
Francis Williams
@frncswllms
Looking for summer internship working at the cutting edge of 3D reconstruction, 3D generation, Neural Fields and more? Our group (nv-tlabs.github.io) is looking for motivated PhD students. DM me with your CV and webpage to to apply!
-
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
We have some openings for PhD internship positions at NVIDIA Toronto AI Lab (nv-tlabs.github.io). If you are interested in doing an internship and working on 3D generative models/reconstruction/neural fields, please DM me with a CV!
-
Sanja Fidler reposted
maggie_albrecht
@maggie_albrecht
Improving Diffusion Models as an Alternative To GANs, Part 1 | NVIDIA Technical Blog developer.nvidia.com/blog/improving-diffusion-models-as-a…
-
Sanja Fidler reposted
Amlan Kar
@amlankar95
We're looking for motivated research interns to join us at NVIDIA Toronto (nv-tlabs.github.io). Shoot me an e-mail if you're interested in HD-mapping or scenario/scene synthesis for autonomous driving! Happy to refer to other team members if interests align. DMs open!
-
Sanja Fidler reposted
Zan Gojcic
@ZGojcic
We have several PhD research internship positions open in our group at NVIDIA! If you are interested in working on neural fields, neural reconstruction or 3D generative models feel to reach out per DM/email. Check also other cool projects from our team at nv-tlabs.github.io
-
Sanja Fidler reposted
Jun Gao
@JunGao33210520
Our group, #nvidia Toronto AI Lab is looking for motivated interns! Email me if you are interested in 3D generation (or reconstruction) for objects (or scenes), or text to 3D!! If you have interests that are related to other people in our group, I'm happy to refer as well!
-
Sanja Fidler reposted
Two Minute Papers
@twominutepapers
New Video - NVIDIA’s New AI: Generating 3D Worlds!
youtu.be/shy51E-MU8Y
#nvidia #ai #rtx -
Sanja Fidler reposted
Masha Shugrina
@_shumash
While some worry AI is taking over #digitalart, we're proud to present Neural Brushstroke Engine #neube that uses #ai to bring new drawing tools to the human #digitalartist. From #nvidia #ai @FidlerSanja at #siggraphasia2022. nv-tlabs.github.io/brushstroke_engine/
-
Sanja Fidler reposted
Jun Gao
@JunGao33210520
I will present this work in person @NeurIPSConf! Come by our poster session on Tue 11am, Hall J, #926 (Yes, it’s tmr!!!)
Please also DM me if you want to chat about the 3D generative model, 3d content creation in general, or just want to catchup!
Jun Gao
@JunGao33210520
Excited to share our #neurips2022 @NVIDIAAI work GET3D, a generative model that directly produces explicit textured 3D meshes with complex topology, rich geometric details, and high fidelity textures. #3d
Project page: nv-tlabs.github.io/GET3D/ -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I’m extremely proud of the @Waabi_ai team for all the amazing work that has gone into making Waabi Driver. This solution marks the next generation of autonomous trucking. I’m so excited to share this with you all!
Waabi
@Waabi_ai
We are extremely excited to unveil the Waabi Driver: the next gen of autonomous trucking technology. Powered by our AI-first approach, the Waabi Driver is safe, scalable, and ready for trucking now. Learn more here: waabi.ai/introducing-the-waabi-driver/
-
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
🦁 LION: Latent Point Diffusion Models for 3D Shape Generation 🦁
nv-tlabs.github.io/LION/
Excited to share our #neurips2022 work on 3D shape generation from direct training on geometry data!
w/ @xiaohuizeng0, @ArashVahdat, @frncswllms, @ZGojcic, @orlitany, @FidlerSanja
(1/n) -
Sanja Fidler reposted
Jose M. Alvarez
@ALVAREZ_JOSEM
87% of AI projects fail to hit production, 96% have data problems & 51% underestimate how much data they need.
Our #neurips2022 paper proposes learning & optimizing data collection!
Sanja Fidler Rafid Mahmood James Lucas and Marc-Law nv-tlabs.github.io/LearnOptimizeCollect/ -
Sanja Fidler reposted
AK
@_akhaliq
LION: Latent Point Diffusion Models for 3D Shape Generation
abs: arxiv.org/abs/2210.06978
project page: nv-tlabs.github.io/LION/ -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
The #ai Animal Explorer Extension lets creators quickly prototype unique #3d animal meshes using probabilistic diffusion models trained on point clouds of animal parts. 🦌
nvda.ws/3CPM0CM -
Sanja Fidler reposted
Masha Shugrina
@_shumash
Excited to share an #ai-powered 3D tool from my group within NVIDIA Toronto AI Lab @NVIDIAAI, available in NVIDIA Omniverse today. Kudos Anita Hu and others on the launch! docs.omniverse.nvidia.com/prod_extensions/prod_extensions…
NVIDIA Omniverse
@nvidiaomniverse
The #ai Animal Explorer Extension lets creators quickly prototype unique #3d animal meshes using probabilistic diffusion models trained on point clouds of animal parts. 🦌
nvda.ws/3CPM0CM -
Sanja Fidler reposted
Francis Williams
@frncswllms
Need to compute Earth Mover's distances between large point clouds in PyTorch? Look no further than: github.com/fwilliams/scalable-pytorch-sinkhorn,
an easy-to-use, scalable implementation of Sinkhorn Distances, which can be differentiated and runs on the GPU! -
Sanja Fidler reposted
Rafid Mahmood
@RafidRMahmood
1/8) 87% of AI projects fail to hit production, 96% have data problems & 51% underestimate how much data they need.
Our #neurips2022 paper proposes learning & optimizing data collection!
w/ @james_r_lucas @ALVAREZ_JOSEM @FidlerSanja @MarcTLaw
Webpage: nv-tlabs.github.io/LearnOptimizeCollect/ -
Sanja Fidler reposted
Jun Gao
@JunGao33210520
The source code for training and inference has been released!! github.com/nv-tlabs/GET3D
Jun Gao
@JunGao33210520
Excited to share our #neurips2022 @NVIDIAAI work GET3D, a generative model that directly produces explicit textured 3D meshes with complex topology, rich geometric details, and high fidelity textures. #3d
Project page: nv-tlabs.github.io/GET3D/ -
Sanja Fidler reposted
Dima Damen
@dimadamen
Take Egocentric vision to the pixel-level!
Thanks to the spirited @_akhaliq for 📨.
VISOR is now publicly available, with camera-ready ArXiv.
Code and models released yesterday [5 repos], with leaderboards on the challenges expected to open early 2023.
epic-kitchens.github.io/VISOR/
AK
@_akhaliq
EPIC-KITCHENS VISOR Benchmark: Video Segmentations and Object Relations
abs: arxiv.org/abs/2209.13064
project page: epic-kitchens.github.io/VISOR/This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
AK
@_akhaliq
EPIC-KITCHENS VISOR Benchmark: Video Segmentations and Object Relations
abs: arxiv.org/abs/2209.13064
project page: epic-kitchens.github.io/VISOR/This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
NVIDIA GET3D is a generative #ai model that can be used to populate virtual worlds with 3D objects and characters. #gtc22
Learn more: nvda.ws/3UzNS9B -
Sanja Fidler reposted
Jun Gao
@JunGao33210520
Excited to share our #neurips2022 @NVIDIAAI work GET3D, a generative model that directly produces explicit textured 3D meshes with complex topology, rich geometric details, and high fidelity textures. #3d
Project page: nv-tlabs.github.io/GET3D/ -
Sanja Fidler reposted
AK
@_akhaliq
GET3D: A Generative Model of High Quality 3D Textured Shapes Learned from Images
abs: arxiv.org/abs/2209.11163
project page: nv-tlabs.github.io/GET3D/ -
Sanja Fidler reposted
Or Perel
@OrPerel
NeRF latents visualized? Check out our demo: the code for Wisp's SIGGRAPH 2022 session is now published online!
The original session recording is available on www.nvidia.com/en-us/on-demand/session/siggraph2022-sigg2…
#nvidiakaolinwisp #nvidia #neuralfields -
Sanja Fidler reposted
Andrew Tao
@drewtao
So proud of our team to finally reveal DLSS3. We hope you love it. It only took 7 years to get here :). youtu.be/PWcNlRI00jo
-
Sanja Fidler reposted
Vector Institute
@VectorInst
TORAS is a new web-based AI-powered annotation platform co-created by Vector Faculty Member @FidlerSanja It provides a repository system allowing for efficient data management and collaboration b/w users and is equipped with human-in-the-loop AI tools.
aidemos.cs.toronto.edu/annotation-suite/ -
Sanja Fidler reposted
Zian Wang
@zianwang97
Our Toronto Annotation Suite (TORAS) is now public! It is an AI-powered annotation web service, combining years of research and engineering efforts from our lab.
If you’re going to annotate images, it’s easy to take a quick trial and we’d love to know your feedback!
Seung Wook Kim
@seungkim0123
📢[NOW LIVE] Toronto Annotation Suite (TORAS) is available to the public!
TORAS is a web-based AI-powered data #annotation platform we have been developing at the @UofT for over 4 years. We hope that you find it useful in annotating your data!
More info: aidemos.cs.toronto.edu/toras -
Sanja Fidler
@FidlerSanja
Our Toronto Annotation Suite (TORAS), an AI-powered annotation webservice, is finally public! It has been in development for almost 5 years by my students at @UofT, in collaboration with Abelium. We hope it's useful to anyone wanting to quickly annotate images! Drop us feedback!
Seung Wook Kim
@seungkim0123
📢[NOW LIVE] Toronto Annotation Suite (TORAS) is available to the public!
TORAS is a web-based AI-powered data #annotation platform we have been developing at the @UofT for over 4 years. We hope that you find it useful in annotating your data!
More info: aidemos.cs.toronto.edu/toras
Sanja Fidler
@FidlerSanja
Thanks to @dimadamen and David Fouhey for stress testing TORAS for more than a year -- in annotating Epic Kitchens-VISOR (epic-kitchens.github.io/VISOR/)! And thanks to @researchuoft for the support and sponsorship!
-
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
📢[NOW LIVE] Toronto Annotation Suite (TORAS) is available to the public!
TORAS is a web-based AI-powered data #annotation platform we have been developing at the @UofT for over 4 years. We hope that you find it useful in annotating your data!
More info: aidemos.cs.toronto.edu/toras -
Sanja Fidler reposted
AK
@_akhaliq
Neural Light Field Estimation for Street Scenes with Differentiable Virtual Object Insertion
abs: arxiv.org/abs/2208.09480
project page: nv-tlabs.github.io/outdoor-ar/ -
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
We had a blast at #siggraph2022 last week demoing our new modular open source #neuralfields NeRF library, Kaolin Wisp. In case you missed the session... the video recording of the overview talk and the Wisp tutorial from @OrPerel is available online! www.nvidia.com/en-us/on-demand/session/siggraph2022-sigg2…
-
Sanja Fidler reposted
Masha Shugrina
@_shumash
Excited to share #nvidia Kaolin Wisp Library for #neuralfields #research such as #nerf. Try it today! github.com/NVIDIAGameWorks/kaolin-wisp Kudos @yongyuanxi @OrPerel @Caenorst @FidlerSanja @NVIDIAAI
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
We had a blast at #siggraph2022 last week demoing our new modular open source #neuralfields NeRF library, Kaolin Wisp. In case you missed the session... the video recording of the overview talk and the Wisp tutorial from @OrPerel is available online! www.nvidia.com/en-us/on-demand/session/siggraph2022-sigg2…
-
Sanja Fidler reposted
Dima Damen
@dimadamen
Today, we release EPIC-KITCHENS VISOR: 10M frames of dense segmentations from 271K of manually segmented masks along with 70K of hand-object relations, for long-term reasoning
Watch our 3min VISOR trailer 📽️ [music on]
youtu.be/yGodQAbYW_E
Project: epic-kitchens.github.io/VISOR/
1/N -
Sanja Fidler reposted
Jason Peng
@xbpeng4
We've released the code for ASE, along with pre-trained models and awesome gladiator motion data from @reallusion!
github.com/nv-tlabs/ASE
We will also be presenting this work at SIGGRAPH today, 2:15pm in East Building, Ballroom A/B. -
Sanja Fidler reposted
Or Perel
@OrPerel
kaolin and wisp #siggraph2022 sessions are on Aug 9 🍻
Be sure to register if you're interested in how wisp can simplify your neural field workflows, or want to hear about the latest advancements in kaolin!
#nvidiakaolinwisp #nvidiakaolin
LINK: nvda.ws/3Bs6kda -
Sanja Fidler reposted
NVIDIA Workstation
@NVIDIAworkstatn
Discover the bleeding edge research that enables us to perceive, create, and experience virtual worlds in these #siggraph2022 sessions.
-
Sanja Fidler reposted
NVIDIA
@nvidia
Watch the global premiere of our documentary, Aug. 10 at 10 a.m. PT, to get a behind-the-scenes look at how our creative, engineering, and research teams pushed the limits of NVIDIA HPC, Al, and @NVIDIAOmniverse to create the company keynote. nvda.ws/3QiXDWI #siggraph2022
-
Sanja Fidler reposted
NVIDIA
@nvidia
Glimpse the future of #ai-infused virtual worlds and discover the research and technology that power them at the #siggraph2022 virtual special address, featuring our CEO and senior leaders on Tue, Aug 9, 9:00 a.m. PDT. x.com/i/broadcasts/1mrGmaMzoWWGy
-
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢 Our new library is on GitHub!!
I'm happy to share NVIDIA Kaolin Wisp, a PyTorch library which provides building blocks and interactive apps to build your own complex neural fields. Presented with a nerf snippet of wisp development.
💻: github.com/NVIDIAGameWorks/kaolin-wisp
NVIDIA AI Developer
@NVIDIAAIDev
Are you interested in #nerfs? NVIDIA Kaolin Wisp provides a framework and the building blocks for neural field research. nvda.ws/3Qkg7WT
-
Sanja Fidler reposted
Fuji Tsang Clément
@Caenorst
Super excited to announce our latest addition to the Kaolin suite! Congratz to the team @yongyuanxi @OrPerel @JoeyLitalien @_shumash @FidlerSanja
NVIDIA AI Developer
@NVIDIAAIDev
Are you interested in #nerfs? NVIDIA Kaolin Wisp provides a framework and the building blocks for neural field research. nvda.ws/3Qkg7WT
-
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
This library has been brewing and marinating for the last 2.5 years and includes contributions from amazing collaborators including @OrPerel, @Caenorst, Charles Loop, @JoeyLitalien, Jonathan Tremblay, @_shumash, @FidlerSanja, and hopefully in the future, you!
-
Sanja Fidler reposted
Or Perel
@OrPerel
🎙️Attention Neural Fields researchers!! This one is for you: today we're proud to officially release our research library #nvidiakaolinwisp. Wisp is packed with an interactive renderer and various building blocks for optimizing NeRF / SDF pipelines.
Code: github.com/NVIDIAGameWorks/kaolin-wisp -
Sanja Fidler
@FidlerSanja
Very exciting release of a new SDK in the Kaolin family -- Kaolin Wisp, a Pytorch library to accelerate research on Neural Fields! Awesome work by @yongyuanxi @OrPerel @Caenorst @JoeyLitalien @_shumash! @NVIDIAAI @nvidiaomniverse #3d #nerf
NVIDIA AI Developer
@NVIDIAAIDev
Are you interested in #nerfs? NVIDIA Kaolin Wisp provides a framework and the building blocks for neural field research. nvda.ws/3Qkg7WT
-
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Are you interested in #nerfs? NVIDIA Kaolin Wisp provides a framework and the building blocks for neural field research. nvda.ws/3Qkg7WT
-
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
Diffusion Models are taking over Deep Generative Learning! Like to get started? 🔥
📢 We re-recorded our ~4 hour tutorial on Diffusion Models we presented at @CVPR this year (w/ @RuiqiGao & @ArashVahdat)!
Youtube: www.youtube.com/watch?v=cS6JQpEY9cs
Slides, etc.: cvpr2022-tutorial-diffusion-models.github.io -
Sanja Fidler reposted
Talking Papers Podcast
@talking_papers
New episode ALERT! 💣🎧
The first episode in our #siggrap2022 collection with @HTDerekLiu about his paper:
📰Learning Smooth Neural Functions via Lipschitz Regularization
Co-authors: @frncswllms, @_AlecJacobson, @FidlerSanja, @orlitany
www.itzikbs.com/lipschitz-mlp 🧵 -
Sanja Fidler reposted
Nick Sharp
@nmwsharp
Some very exciting career news:
I've joined @NVIDIA as a scientist to work on the future of 3D geometry, machine learning, and visual computing with @FidlerSanja and many other brilliant folks!
Can't wait to share what we create 🚀(1/2) -
Sanja Fidler reposted
-
Sanja Fidler reposted
Rafid Mahmood
@RafidRMahmood
(1/7) Our #cvpr2022 paper “How Much More Data Do I Need? Estimating Requirements For Downstream Tasks” explores data collection practices!
w/ @james_r_lucas @davidjesusacu @lidaiqing @PhilionJonah @ALVAREZ_JOSEM @ZhidingYu @FidlerSanja @MarcTLaw
Webpage: nv-tlabs.github.io/estimatingrequirements/ -
Sanja Fidler
@FidlerSanja
Synthetic data generation now available in #omniversecode!
NVIDIA Omniverse
@nvidiaomniverse
Omniverse Replicator SDK is NOW available in #omniversecode. 🙌
Learn how developers, researchers, & engineers can use #omniversereplicator to bootstrap & improve performance of existing #deeplearning perception models with large-scale photorealistic #syntheticdatageneration. -
Sanja Fidler reposted
Amulya Vishwanath
@amulyavy
NVIDIA Academic Hardware Grant Program Deadline 📢
Closes: July 1, 2022 6:00 PM PT
Award Decisions Sent By: August 26th, 2022. Application Info & FAQ 👇🏽 -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Jazz is all about improvisation — and NVIDIA is paying tribute to the genre with #ai research that could one day enable graphics creators to improvise with 3D objects created in the time it takes to hold a jam session. nvda.ws/3n2V8v9 #nerf #cvpr2022
-
Sanja Fidler reposted
maggie_albrecht
@maggie_albrecht
AI in the Big Easy: NVIDIA Research Lets Content Creators Improvise With 3D Objects share.nvidia.com/Article/Redirect/d8992768-3c7b-4939-bcfa…
-
Sanja Fidler reposted
Ben Poole
@poolio
Wow this dataset will be amazingly useful for folks in vision science! Spent too many hours crafting "natural" stimuli in my PhD...
Daiqing Li
@lidaiqing
Excited to share BigDatasetGAN @CVPR! We are able to synthesize ImageNet with pixel-wise labels using as few as 5 annotations per class! Joint work w/ @HuanLing6 @seungkim0123 @karsten_kreis
@FidlerSanja @abtorralba
@NVIDIAAI
Dataset/code are released! github.com/nv-tlabs/bigdatasetgan_code
AK
@_akhaliq
BigDatasetGAN: Synthesizing ImageNet with Pixel-wise Annotations
abs: arxiv.org/abs/2201.04684
project page: nv-tlabs.github.io/big-datasetgan/ -
Sanja Fidler reposted
Despoina Paschalidou
@paschalidoud_1
Happening now #127a! Another great work by @NVIDIAAI at @CVPR!!! Big shout out to @davrempe @orlitany @FidlerSanja and the rest of the team for this great paper! Make sure to stop by!!
Davis Rempe
@davrempe
Come by poster 127 this morning at #cvpr to learn how we can automatically generate accident scenarios to improve AV planners!
Video: youtu.be/j3gS0TY3iOw
Project: nv-tlabs.github.io/STRIVE/ -
Sanja Fidler reposted
Francis Williams
@frncswllms
(1/N) 🎷🎷🎷 Check out our work on Neural Kernel Fields (NKF) at #cvpr2022 at Halls B2-C! We present a new type of #neuralfields based on learned kernels. NKF pushes the state of the art in shape reconstruction from points and shows unprecedented generalization ability:
-
Sanja Fidler reposted
Davis Rempe
@davrempe
Come by poster 127 this morning at #cvpr to learn how we can automatically generate accident scenarios to improve AV planners!
Video: youtu.be/j3gS0TY3iOw
Project: nv-tlabs.github.io/STRIVE/ -
Sanja Fidler reposted
AK
@_akhaliq
BigDatasetGAN: Synthesizing ImageNet with Pixel-wise Annotations
abs: arxiv.org/abs/2201.04684
project page: nv-tlabs.github.io/big-datasetgan/ -
Sanja Fidler reposted
Daiqing Li
@lidaiqing
Excited to share BigDatasetGAN @CVPR! We are able to synthesize ImageNet with pixel-wise labels using as few as 5 annotations per class! Joint work w/ @HuanLing6 @seungkim0123 @karsten_kreis
@FidlerSanja @abtorralba
@NVIDIAAI
Dataset/code are released! github.com/nv-tlabs/bigdatasetgan_code
AK
@_akhaliq
BigDatasetGAN: Synthesizing ImageNet with Pixel-wise Annotations
abs: arxiv.org/abs/2201.04684
project page: nv-tlabs.github.io/big-datasetgan/ -
Sanja Fidler reposted
Aaron Lefohn
@aaronlefohn
Jazz instruments from photos?
Yes. We used our new CVPR paper's differentiable rendering pipeline to reconstruct geometry, materials, and lighting from photos.
More details in our CVPR paper w/@FidlerSanja's team, presented tomorrow. (1/3)
www.youtube.com/watch?v=W1VhDtoTzA4 -
Sanja Fidler reposted
Matan
@matanatzmon
Happy to share our #cvpr2022 work on Frame Averaging for Equivariant Shape Space Learning.
paper: arxiv.org/abs/2112.01741
project: nv-tlabs.github.io/equivariant/
video: www.youtube.com/watch?v=Lft6r5oVyXM
Joint work with @luminohope, @FidlerSanja , @sameh_khamis, @lipmanya -
Sanja Fidler reposted
Christian Wolf (🦋🦋🦋)
@chriswolfvision
Woa... a new large-scale dataset for video object segmentation, first person video, by @dimadamen's group.
-
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢 Our new SIGGRAPH 2022 paper is on arXiv!!
I'm happy to share Variable Bitrate Neural Fields, which lets you download ultra compact & high fidelity & fast NeRFs and neural fields at ⚡ blazing speeds with progressive LOD!
🌐: nv-tlabs.github.io/vqad/
🎥: www.youtube.com/watch?v=Lh0CoTRNFBA -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
Diffusion Models are revolutionizing Deep Generative Modeling!
📢 At @CVPR, together with @RuiqiGao and @ArashVahdat, I will be presenting a tutorial on Diffusion Models: cvpr2022-tutorial-diffusion-models.github.io
Find a Diffusion Model-generated impression of what to expect below 😀
#cvpr2022 -
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
Excited to share Polymorphic-GAN (@ CVPR2022 as an oral)! We introduce a GAN model that generates aligned samples across multiple domains. Joint work w/
@karsten_kreis @lidaiqing @abtorralba @FidlerSanja @NVIDIAAI
Webpage: nv-tlabs.github.io/PMGAN
Paper: arxiv.org/abs/2206.02903 -
Sanja Fidler reposted
Two Minute Papers
@twominutepapers
NVIDIA’s New AI Grows Objects Out Of Nothing! 🤖
▶️Full video (ours): youtu.be/5j8I7V6blqM
📜Source paper: nvlabs.github.io/nvdiffrec/
#nvidia #ai -
Sanja Fidler reposted
Hsueh-Ti Derek Liu
@HTDerekLiu
Happy to share our #siggraph2022 paper for encouraging smooth neural fields, useful for geometric tasks e.g. interpolation. Thanks to amazing collaborators: @frncswllms @_AlecJacobson @FidlerSanja @orlitany. (1/n)
Project page: nv-tlabs.github.io/lip-mlp/ -
Sanja Fidler
@FidlerSanja
Check out our SIGGRAPH paper with @xbpeng4 on character animation! @nvidiaomniverse @NVIDIAAI
Jason Peng
@xbpeng4
Our paper on learning reusable skill embeddings for simulated characters has been accepted into #siggraph2022!
Paper: nv-tlabs.github.io/ASE/
Thanks to all my collaborators: Kelly Guo, Lina Halper, @svlevine, @FidlerSanja, and all the folks at NVIDIA -
Sanja Fidler reposted
Jason Peng
@xbpeng4
Our paper on learning reusable skill embeddings for simulated characters has been accepted into #siggraph2022!
Paper: nv-tlabs.github.io/ASE/
Thanks to all my collaborators: Kelly Guo, Lina Halper, @svlevine, @FidlerSanja, and all the folks at NVIDIA -
Sanja Fidler reposted
NVIDIA Workstation
@NVIDIAworkstatn
From neural rendering to holography, NVIDIA will be sharing a record 16 research papers on advancements in graphics this year at @siggraph. Get a sneak peek in our blog.
#siggraph2022 nvda.ws/3KArI0E -
Sanja Fidler reposted
Two Minute Papers
@twominutepapers
NVIDIA's New AI: Next Level Image Editing! 👌
▶️Full video (ours): www.youtube.com/watch?v=cS4jCvzey-4
📜Source paper: nv-tlabs.github.io/editGAN/
#nvidia #ai -
Sanja Fidler reposted
Zhiqin Chen
@ZhiqinChen3
(1/4) Pleased to announce our @NVIDIAAI #cvpr2022 paper AUV-Net!
AUV-Net learns aligned UV maps for a set of 3D shapes, allowing us to easily transfer textures between 3D shapes by simply swapping their texture images.
nv-tlabs.github.io/AUV-NET
with @kangxue_yin @FidlerSanja -
Sanja Fidler
@FidlerSanja
Check out our @NVIDIAAI CVPR paper on extracting triangle meshes, materials, and lighting directly from images! Code is released.
yosl ber
@yoslber
Want to extract triangle meshes, materials, and lighting directly from images?
Paper, video & code: nvlabs.github.io/nvdiffrec/ CVPR 2022 (Oral).
Nice collaboration between @aaronlefohn and @FidlerSanja teams at @nvidia.
AK
@_akhaliq
Extracting Triangular 3D Models, Materials, and Lighting From Images
abs: arxiv.org/abs/2111.12503
present an efficient method for joint optimization of topology, materials and lighting from multi-view image observations -
Sanja Fidler reposted
yosl ber
@yoslber
Want to extract triangle meshes, materials, and lighting directly from images?
Paper, video & code: nvlabs.github.io/nvdiffrec/ CVPR 2022 (Oral).
Nice collaboration between @aaronlefohn and @FidlerSanja teams at @nvidia.
AK
@_akhaliq
Extracting Triangular 3D Models, Materials, and Lighting From Images
abs: arxiv.org/abs/2111.12503
present an efficient method for joint optimization of topology, materials and lighting from multi-view image observations -
Sanja Fidler reposted
Lina (혜원)
@nikolobin
See you in the Omniverse at #gtc22 😎
Registration for my session is open and free to attend. #nvidiaomniverse @NVIDIAGameDev @nvidiaomniverse
nvda.ws/3D7836j
A sneak preview of Animation Retargeter coming to Omniverse. -
Sanja Fidler reposted
NVIDIA DRIVE
@NVIDIADRIVE
Learn about the breakthroughs from NVIDIA Research and @nvidiaomniverse that enable developers to reconstruct scenarios in simulation, accelerating AV development: nvda.ws/3IGnfZu #gtc22
-
Sanja Fidler
@FidlerSanja
Really proud of the team and collaborators at @nvidia to deliver this great work on driving simulation!
Or Litany
@orlitany
Check out our recent progress on generating challenging scenarios and re-enacting pre-recorded drives for closed-loop simulation #gtc22
NVIDIA DRIVE
@NVIDIADRIVE
Get a first look at novel #ai tools for scenario reconstruction in @nvidiaomniverse — a key building block for #autonomousvehicle training, testing and validation: nvda.ws/3JzE9dj #gtc22
-
Sanja Fidler reposted
Or Litany
@orlitany
Check out our recent progress on generating challenging scenarios and re-enacting pre-recorded drives for closed-loop simulation #gtc22
NVIDIA DRIVE
@NVIDIADRIVE
Get a first look at novel #ai tools for scenario reconstruction in @nvidiaomniverse — a key building block for #autonomousvehicle training, testing and validation: nvda.ws/3JzE9dj #gtc22
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Super excited to welcome the amazing Vivian Sun to our team as Chief Commercial Officer!
Waabi
@Waabi_ai
We're excited to welcome Vivian Sun to the team as Chief Commercial Officer! Our journey to rapidly & safely commercialize self-driving vehicles starts with logistics—Vivian will be instrumental in building partnerships & introducing our tech to the market
waabi.ai/vivian-sun-cco/ -
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
Just released: #nvidiakaolin 0.10.0!
New features of the launch include Volumetric rendering, additional features for tetrahedral meshes, and new support for materials with @OpenUSD importation.
Don’t forget to test our two new tutorials on 👉 nvda.ws/3sRiO87 -
Sanja Fidler reposted
Arash Vahdat
@ArashVahdat
🎉 Our tutorial proposal on 𝗗𝗲𝗻𝗼𝗶𝘀𝗶𝗻𝗴 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻-𝗯𝗮𝘀𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗠𝗼𝗱𝗲𝗹𝗶𝗻𝗴 has been accepted to #cvpr2022. Stay tuned for some cool lectures from @karsten_kreis, @RuiqiGao and me at @CVPR 2022.
-
Sanja Fidler reposted
Huan Ling
@HuanLing6
We are excited to announce that code and tool for our
EditGAN are released!
Code: github.com/nv-tlabs/editGAN_release
w/ @karsten_kreis , @lidaiqing , @seungkim0123 , @abtorralba , @FidlerSanja
@NVIDIAAI
Huan Ling
@HuanLing6
(1/6) Excited to share our #nvidiaai #neurips2021 paper “EditGAN: High-Precision Semantic Image Editing” (nv-tlabs.github.io/editGAN/), achieving an unprecedented level of detail in GAN-based image editing!
w/ @karsten_kreis , @lidaiqing , @seungkim0123 , @abtorralba , @FidlerSanja -
Sanja Fidler reposted
Cinjon Resnick
@cinjoncin
(re)announcing Causal Scene BERT. Understand in what regions an object detector will have subpar scores BEFORE it goes into the wild. We do this by using masked language models to take causal interventions over visual scenes --> arxiv.org/abs/2202.03651.
Kyunghyun Cho
@kchonyc
i never got to tweet about it earlier, but it's always fun to use an idea (here, BERT for plausible perturbation) from one domain (here NLP; arxiv.org/abs/2009.10195) to another domain (here CV for self-driving; openaccess.thecvf.com/content/ICCV2021W/AVVision/html/Res…) which was spearheaded by @cinjoncin
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I am so proud of @Waabi_ai team and everything we have achieved only 6 months on from our launch. Waabi World is the next generation of simulation technology and will be the key to unlocking self-driving at scale. We’re so excited about what’s to come!
waabi.ai/waabi-world/
Waabi
@Waabi_ai
Today we are unveiling our revolutionary simulator—Waabi World. #waabiworld is the most scalable, highest fidelity closed-loop simulator ever and the key to rapidly and safely commercializing self-driving technology. waabi.ai/waabi-world/
-
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
@ArashVahdat set the bar very high today with his excellent and super insightful talk!
I am very excited to present our (w/ @timudk) other work along the same line of research on critically-damped Langevin diffusion (nv-tlabs.github.io/CLD-SGM/) in 2 weeks! 🙂 -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
2022 is a critical year for the self-driving industry—I’m excited to share a few thoughts on what the industry needs to do in order to make the dream of commercial self-driving cars a reality
www.linkedin.com/pulse/how-we-get-self-driving-back-track…
#autonomousdriving #selfdrivingcars -
Sanja Fidler reposted
NVIDIA
@nvidia
Thanks to our employees, NVIDIA is at No. 1 on @Glassdoor's "Best Places to Work" list of large U.S. companies, according to newly released rankings. nvda.ws/3zU3rz2 #glassdoorbptw
-
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
JUST ANNOUNCED: #nvidiaomniverse is now available free to millions of creators to accelerate their #3d workflows with NVIDIA's core rendering, physics, and #ai technologies.
Read more. -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
No need for expensive motion-capture hardware.
#ai researchers from @NVIDIA developed a new #physics-based method to estimate and synthesize human motion models from 2D videos. Learn more: nvda.ws/324ov9p -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
📢 Score-based Generative Modeling with Critically-Damped Langevin Diffusion!
nv-tlabs.github.io/CLD-SGM/
We propose a novel diffusion using auxiliary velocity variables for more efficient denoising and higher quality generative models.
w/ the amazing @timudk & @ArashVahdat!
(1/n) -
Sanja Fidler reposted
Tim Dockhorn
@timudk
Super excited to finally reveal what I have been working on with @karsten_kreis and @ArashVahdat during my internship at nvidia.
I am also very happy to announce that I will stay on this amazing team led by @FidlerSanja as an intern, and push score-based models even further.
Karsten Kreis
@karsten_kreis
📢 Score-based Generative Modeling with Critically-Damped Langevin Diffusion!
nv-tlabs.github.io/CLD-SGM/
We propose a novel diffusion using auxiliary velocity variables for more efficient denoising and higher quality generative models.
w/ the amazing @timudk & @ArashVahdat!
(1/n) -
Sanja Fidler reposted
James Lucas
@james_r_lucas
I'm super excited to share that I'll be joining @nvidia as a research scientist early next year!
I feel very fortunate that I can continue to work with such an amazing team, led by @FidlerSanja. I'll be moving back to the UK (!) and I can't wait to start this next chapter. -
Sanja Fidler reposted
Tianchang Shen
@TianchangS
Check out our new @NVIDIAAI #neurips2021 paper DMTet (shorturl.at/fpHP1) which synthesizes high-resolution 3D shapes from coarse voxels or noisy point cloud!
w/@JunGao33210520 @kangxue_yin @liu_mingyu @FidlerSanja
Today Dec 10 11:30am - 1 pm EST. Come and talk to us! -
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
This was an exciting project - I continue to be amazed by the capabilities of modern deep generative models! If you are interested in working with us on generative models and their applications, please reach out. We are looking for exceptional interns at NVIDIA's Toronto AI Lab.
Huan Ling
@HuanLing6
(1/6) Excited to share our #nvidiaai #neurips2021 paper “EditGAN: High-Precision Semantic Image Editing” (nv-tlabs.github.io/editGAN/), achieving an unprecedented level of detail in GAN-based image editing!
w/ @karsten_kreis , @lidaiqing , @seungkim0123 , @abtorralba , @FidlerSanja -
Sanja Fidler reposted
Huan Ling
@HuanLing6
(1/6) Excited to share our #nvidiaai #neurips2021 paper “EditGAN: High-Precision Semantic Image Editing” (nv-tlabs.github.io/editGAN/), achieving an unprecedented level of detail in GAN-based image editing!
w/ @karsten_kreis , @lidaiqing , @seungkim0123 , @abtorralba , @FidlerSanja -
Sanja Fidler reposted
Wenzheng Chen
@ChenWenzheng
(1/4) Excited to share our new @NVIDIAAI #neurips2021 work DIB-R++.
nv-tlabs.github.io/DIBRPlus/
@JoeyLitalien @JunGao33210520 @zianwang97 @Caenorst @sameh_khamis @orlitany @FidlerSanja
We jointly predict 3D shape, texture, lighting as well as material from monocular photographs! -
Sanja Fidler reposted
David Acuna
@davidjesusacu
(1/4) Excited to share our new #neurips2021 paper
Towards Optimal Strategies for Training Self-Driving Perception Models in Simulation
nv-tlabs.github.io/simulation-strategies/
Tmr Dec 7, 11:30-1pm EST. Come and talk to us!
This is joint work with @PhilionJonah, @FidlerSanja at @NVIDIAAI -
Sanja Fidler reposted
Despoina Paschalidou
@paschalidoud_1
📢Check out our new @NeurIPSConf paper in collaboration with @NVIDIAAI!!!
ATISS is a novel transformer architecture that advances the state-of-the-art of scene synthesis while enabling a number of interactive applications with versatile user input.
nv-tlabs.github.io/ATISS -
Sanja Fidler reposted
U of T Department of Computer Science
@UofTCompSci
Join us at @UofTCompSci! We are now hiring for tenure-stream faculty positions in multiple areas.
We will start reviewing applications on December 6, and we will give full consideration to all applications submitted by January 10.
Get the details: uoft.me/7jH -
Sanja Fidler
@FidlerSanja
Code for this project is now released: nv-tlabs.github.io/fDAL/ !
@davidjesusacu @MarcTLaw @NVIDIAAI
David Acuna
@davidjesusacu
Excited to share fDAL #icml2021
We connect theory and algorithms for domain adaptation using a novel f-divergence based bound. fDAL explains existing algorithms, shows their bottlenecks and how to fix them!
arxiv.org/abs/2106.11344
w/ @gjzhang1 @MarcTLaw @FidlerSanja @NVIDIAAI -
Sanja Fidler reposted
AK
@_akhaliq
Deep Marching Tetrahedra: a Hybrid Representation
for High-Resolution 3D Shape Synthesis
abs: arxiv.org/abs/2111.04276
project page: nv-tlabs.github.io/DMTet/ -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
#nvidiaomniverse Avatar, a technology platform to develop interactive #ai avatars that respond to speech and facial cues, can understand multiple languages, and are able to make smart recommendations was just unveiled at #gtc21.
-
Sanja Fidler
@FidlerSanja
Great opportunity to work with Dima on video understanding! I feel exactly like the Minions in the tweet!
Dima Damen
@dimadamen
📢📢3-year postdoc on video understanding in my group
www.bristol.ac.uk/jobs/find/details/?jobId=261315&jobTitl…
DM or apply directly by 5 Dec
U can choose your topic-of-focus within multi-modal video understanding
We are *proudly* influential but also a cohesive and supportive environment to work within.
Pls r/w -
Sanja Fidler reposted
AK
@_akhaliq
EditGAN: High-Precision Semantic Image Editing
abs: arxiv.org/abs/2111.03186
project page: nv-tlabs.github.io/editGAN/
a novel method for high-quality, high-precision semantic image editing, allowing users to edit images by modifying their highly detailed part segmentation masks -
Sanja Fidler reposted
Animesh Garg
@animesh_garg
Grad Dept at CS has varied research foci! web.cs.toronto.edu/research/areas
Many of us are hiring (not exhaustive!)
@florian_shkurti @igilitschenski @BurgnerKahrs @FidlerSanja @DavidDuvenaud @RogerGrosse @SoloGen @roydanroy @NicolasPapernot @A_Aspuru_Guzik
@VectorInst @UofTRobotics -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Big thanks to @vipalmonga from @WSJ for speaking with me about what we’re building at @Waabi_ai and our plans for commercializing self-driving technology! www.wsj.com/articles/a-former-uber-executive-takes-a-fres…
-
Sanja Fidler reposted
Karsten Kreis
@karsten_kreis
It turns out, when training generative models with strict differential privacy guarantees (a super hard task!), optimal transport is a great framework, outperforming previous GAN-based methods!
#neurips2021 paper together with Tianshi Cao, Alex Bie, @ArashVahdat and @FidlerSanja
AK
@_akhaliq
Don’t Generate Me: Training Differentially Private Generative Models with Sinkhorn Divergence
abs: arxiv.org/abs/2111.01177 -
Sanja Fidler reposted
NVIDIA GTC
@NVIDIAGTC
What will Jensen Huang's #gtc21 keynote reveal? Be sure to tune in on November 9. nvda.ws/2XXLsJn
-
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
Code for "DriveGAN: Towards a Controllable high-Quality Neural Simulation" (CVPR 2021, Oral) is now released!
Webpage: nv-tlabs.github.io/DriveGAN/
Code: github.com/nv-tlabs/DriveGAN_code
Paper: arxiv.org/abs/2104.15060
Joint work with
@PhilionJonah @abtorralba @FidlerSanja @NVIDIAAI -
Sanja Fidler reposted
Srinath Sridhar
@drsrinathsridha
We are organizing the Second #3dreps workshop at @ICCV_2021 to bring together researchers working on learned 3D representations. We have an amazing lineup of invited speakers and posters. Please join us tomorrow (Oct 17)! Full schedule here: ivl.cs.brown.edu/3DReps
-
Sanja Fidler reposted
Florian Shkurti
@florian_shkurti
Kevin Xie will be presenting on physics-based human pose estimation and synthesis from videos at ICCV@ 5pm EDT today in the poster session. Drop by and say hi if you like contact invariant opt. Joint work w/ @umariqb @TingwuWang @kellyg1110 @FidlerSanja
AK
@_akhaliq
Physics-based Human Motion Estimation and Synthesis from Videos
pdf: arxiv.org/pdf/2109.09913.pdf
abs: arxiv.org/abs/2109.09913
introduced a new framework for training motion synthesis models from raw video pose estimations without making use of motion capture data -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Today, together with my collaborators Andreas Geiger, Philip Lenz and Christoph Stiller, I was awarded the 2021 Everingham Prize for KITTI at #iccv21, which enabled many breakthroughs in #selfdrivingcars. Thank you, truly an honor!
-
Sanja Fidler reposted
Zian Wang
@zianwang97
Excited to share our #iccv2021 paper “Learning Indoor Inverse Rendering with 3D Spatially-Varying Lighting”! It handles 3D lighting, disentangles spatially-varying effects, and benefits AR applications.
w/ @PhilionJonah @FidlerSanja @jankautz @NVIDIAAI
nv-tlabs.github.io/inverse-rendering-3d-lighting/ -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I’m excited to announce that I will be speaking at this year’s @WSJ Tech Live event. Looking forward to speaking with @timkhiggins on Waabi’s #ai first approach to self-driving cars.
Waabi
@Waabi_ai
CEO @RaquelUrtasun will be speaking at @WSJ Tech Live event this month! We invite you to tune in on 10/18 at 12:30 PM ET to listen to her talk on Waabi’s #ai first approach to #selfdriving #cars ➡️ techlive.wsj.com/person/raquel-urtasun/ #wsjtechlive
-
Sanja Fidler reposted
AK
@_akhaliq
Physics-based Human Motion Estimation and Synthesis from Videos
pdf: arxiv.org/pdf/2109.09913.pdf
abs: arxiv.org/abs/2109.09913
introduced a new framework for training motion synthesis models from raw video pose estimations without making use of motion capture data -
Sanja Fidler reposted
Umar Iqbal
@UmarIqb
Excited to share our #iccv2021 paper:
Physics-based Human Motion Estimation and Synthesis from Videos
pdf: arxiv.org/pdf/2109.09913.pdf
abs: arxiv.org/abs/2109.09913
Project page: nv-tlabs.github.io/physics-pose-estimation-project-page/
with Kevin Xie, @TingwuWang @FidlerSanja, and Florian Shkurti
AK
@_akhaliq
Physics-based Human Motion Estimation and Synthesis from Videos
pdf: arxiv.org/pdf/2109.09913.pdf
abs: arxiv.org/abs/2109.09913
introduced a new framework for training motion synthesis models from raw video pose estimations without making use of motion capture data -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Research internships are now available
@Waabi_ai. All year long, with duration of 3-12 months. Available in both Canada as well as US. Join the team at the forefront of innovation in #selfdrivingcars !
Apply: jobs.lever.co/waabi/ad3c77da-5607-4c4f-8784-8407478cbace -
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢Here's something cool we've been working on: efficient sparse octree primitives for 3D deep learning with PyTorch. This enables training NGLOD 3x faster with 30x less memory, which also means it can fit 30x more complex models! Code available here: github.com/nv-tlabs/nglod
NVIDIA AI Developer
@NVIDIAAIDev
Just released: 3D DL researchers can build on the latest algorithms to simplify and accelerate workflows using Kaolin PyTorch Library. Learn more: nvda.ws/3zKvgsD
-
Sanja Fidler
@FidlerSanja
Check out the new release of Kaolin! It includes a sparse octree-based acceleration data structure, with highly efficient convolution and ray tracing capabilities. SPCs are useful for scaling up and accelerating neural implicit representations! #3d #deeplearning
NVIDIA AI Developer
@NVIDIAAIDev
Just released: 3D DL researchers can build on the latest algorithms to simplify and accelerate workflows using Kaolin PyTorch Library. Learn more: nvda.ws/3zKvgsD
-
Sanja Fidler reposted
Kangxue Yin
@kangxue_yin
Excited to share #3dstylenet, a Neural Style Transfer method for 3D Shapes! The work is accepted to #iccv2021 as an oral presentation.
arXiv: arxiv.org/abs/2108.12958
Project page: nv-tlabs.github.io/3DStyleNet/
with @JunGao33210520 @_shumash @sameh_khamis @FidlerSanja
@NVIDIAAI -
Sanja Fidler reposted
AK
@_akhaliq
3DStyleNet: Creating 3D Shapes with Geometric and Texture Style Variations
pdf: arxiv.org/pdf/2108.12958.pdf
abs: arxiv.org/abs/2108.12958
project page: nv-tlabs.github.io/3DStyleNet/ -
Sanja Fidler reposted
-
Sanja Fidler reposted
Aaron Lefohn
@aaronlefohn
Computer science Ph.D. students, apply for the NVIDIA Graduate Fellowship. Applications are now open, and the deadline is September 8th.
www.nvidia.com/en-us/research/graduate-fellowships/ -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
#nvidiaresearch comes to #nvidiaomniverse with the new GANverse3D + Image2Car extension, the first entry in the Omniverse #ai Toy Box.
See an artist use it to create #3d cars from 2D images, and use real-time #physx5 to stack & explode the cars. nvda.ws/3jvfgDR -
Sanja Fidler
@FidlerSanja
Check out GANverse3D, that converts images of cars into 3D models, in action!
NVIDIA Omniverse
@nvidiaomniverse
#nvidiaresearch comes to #nvidiaomniverse with the new GANverse3D + Image2Car extension, the first entry in the Omniverse #ai Toy Box.
See an artist use it to create #3d cars from 2D images, and use real-time #physx5 to stack & explode the cars. nvda.ws/3jvfgDR -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
August 10 @ 9AM PDT - Come explore the latest #3d research coming from the @NVIDIA Toronto #ai Lab, including updates to #nvidiakaolin and a new extension for #omniversecreate.
Attend the #siggraph2021 session free with code "NVIDIA21" nvda.ws/3CqBpfn -
Sanja Fidler reposted
Gavriel State
@gavrielstate
Tomorrow at SIGGRAPH I’ll be presenting some of the work that NVIDIA’s Toronto AI team has been doing to create amazing new tools for 3D graphics, alongside my colleagues @_shumash and JF Lafleche.
Plus even more with @FidlerSanja + Richard Kerris here: www.nvidia.com/en-us/events/siggraph/?ncid=so-twit-346272 -
Sanja Fidler reposted
Exploring Slovenia
@ExploringSlo
One of the most overlooked places in Slovenia that everyone should visit at least once. The hills above Krško offer a unique experience of outdoor adventures and incredible cuisine paired with the best local wine. Read about it in my latest BLOG www.exploringslovenia.com/blog/2021/07/04/exquisite-krsko…
-
Sanja Fidler reposted
Exploring Slovenia
@ExploringSlo
As tourists stroll around the deep green Alpine lake Bled, the dramatic backdrop of the Karavanke Mountains lure the other kind – the adventurers at heart. The highest in the 120km-long mountain range is the outstanding Mt Stol. Tours available! More info: www.exploringslovenia.com/tours/stol/
-
Sanja Fidler reposted
U of T Department of Computer Science
@UofTCompSci
We look forward to welcoming @DaveLindell to @UofTCompSci as an assistant professor in July 2022. His research is at the intersection of machine learning, computational imaging, and computer vision. uoft.me/6T3
-
Sanja Fidler reposted
NVIDIA Robotics
@NVIDIARobotics
Join us on July 12th at #rss. This workshop will introduce the end-to-end GPU accelerated training pipeline in #nvidia Isaac Gym, demonstrate #robotics applications, and answer questions in breakout sessions. Register here: nvda.ws/3hzBe7y
#ai #robots #nvidiaisaac -
Sanja Fidler reposted
Bryan Catanzaro
@ctnzr
Today @nvidia launched Cambridge-1, which is now the largest supercomputer in the UK. I expect there will be a lot of research done on this computer at the intersection of HPC, AI, and healthcare. Can’t wait to see what people do with this machine!
www.nvidia.com/en-us/industries/healthcare-life-sciences/… -
Sanja Fidler reposted
NVIDIA
@nvidia
Tune into the virtual inauguration as we unveil transformative research on Cambridge-1, the UK’s most powerful supercomputer. #nvidiacambridge1 x.com/i/broadcasts/1OwxWVVYbXkJQ
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I had a great conversation with @MarwaEldiwiny about AI, #selfdriving, #innovation and my path to found @Waabi_ai.
Marwa ElDiwiny
@MarwaEldiwiny
I enjoyed this conversation with Raquel Urtasun @RaquelUrtasun on the developed self-driving cars at Waabi @Waabi_ai. I hope you enjoy listening #robotics
The episode:
soundcloud.com/ieeeras-softrobotics/raquel-urtasun-waabi-… -
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
Just announced: NVIDIA partners with @GoogleCloud to create the industry’s first #aion5g lab to speed #ai application development for #5g network operators. Learn more.
-
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
Meet Cambridge-1, the UK's fastest supercomputer, designed to solve pressing medical challenges with #ai. On July 7th, watch leading research institutions reveal groundbreaking research planned for #nvidiacambridge1.
-
Sanja Fidler reposted
David Acuna
@davidjesusacu
Excited to share fDAL #icml2021
We connect theory and algorithms for domain adaptation using a novel f-divergence based bound. fDAL explains existing algorithms, shows their bottlenecks and how to fix them!
arxiv.org/abs/2106.11344
w/ @gjzhang1 @MarcTLaw @FidlerSanja @NVIDIAAI -
Sanja Fidler
@FidlerSanja
Domain adaptation is a problem that appears very frequently in ML pipelines. Check out latest @davidjesusacu's work which gets state-of-the-art on a few computer vision and NLP problems and it's a generic method you can easily plug&play! @NVIDIAAI
David Acuna
@davidjesusacu
Excited to share fDAL #icml2021
We connect theory and algorithms for domain adaptation using a novel f-divergence based bound. fDAL explains existing algorithms, shows their bottlenecks and how to fix them!
arxiv.org/abs/2106.11344
w/ @gjzhang1 @MarcTLaw @FidlerSanja @NVIDIAAI -
Sanja Fidler reposted
Renjie Liao
@lrjconan
Happy to share our UAI 2021 paper on learning discrete VAEs for iterative image generation, like human drawing! The key is a non-parametric prior over a sequence of discrete decisions (whether, where, & what-to-draw).
Paper: arxiv.org/abs/2106.13435
Code: github.com/ZENGXH/NPDRAW -
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
This #ai model can take one 2D image of a person and make a standard video conference more engaging and fun by synthesizing a digital avatar in real time, and making AI eye contact. NVIDIA Vid2Vid Cameo steals the show! Learn more: nvda.ws/3vVxylT #cvpr2021
-
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
Join NVIDIA researchers today for a presentation on "DatasetGAN: Efficient Labeled Data Factory With Minimal Human Effort" at #cvpr2021. #datasetgan is able to reduce the need for annotation 100x times. Learn more: nvda.ws/3xO1OAA
-
Sanja Fidler
@FidlerSanja
Code and data for DatasetGAN released! Please also drop by Huan’s poster session!
Huan Ling
@HuanLing6
This is code and data for our
@CVPR'21 oral paper DatasetGAN! github.com/nv-tlabs/datasetGAN_release/
Project Page: nv-tlabs.github.io/datasetGAN
CVPR Poster Session, Wednesday, June 23, 2021 10:00 PM – 12:30 AM
@YuxuanZ84175540 @FidlerSanja @kangxue_yin @JunGao33210520 @NVIDIAAI -
Sanja Fidler reposted
Huan Ling
@HuanLing6
This is code and data for our
@CVPR'21 oral paper DatasetGAN! github.com/nv-tlabs/datasetGAN_release/
Project Page: nv-tlabs.github.io/datasetGAN
CVPR Poster Session, Wednesday, June 23, 2021 10:00 PM – 12:30 AM
@YuxuanZ84175540 @FidlerSanja @kangxue_yin @JunGao33210520 @NVIDIAAI -
Sanja Fidler reposted
Daiqing Li
@lidaiqing
We are presenting our work on SemanticGAN at @CVPR Please join us! Released code is now also available! @nvidiaai @junlinnnnnnn @karsten_kreis, Antonio, @FidlerSanja
presentation at CVPR: Wed 6-8:30 AM EST
project page: nv-tlabs.github.io/semanticGAN/
code: github.com/nv-tlabs/semanticGAN_code -
Sanja Fidler
@FidlerSanja
Come check out our DriveGAN paper today at CVPR!
NVIDIA AI
@NVIDIAAI
Today during #cvpr2021 see featured paper co-authored by Director of #ai at #nvidia @FidlerSanja "DriveGAN: Towards a Controllable High-Quality Neural Simulation." Learn more:
-
Sanja Fidler reposted
NVIDIA Robotics
@NVIDIARobotics
Announcing NVIDIA #isaac Sim, now on @NVIDIAOmniverse. Be one of the first to experience the open beta to build, test, and manage AI-based robots in a virtual, physically accurate, collaborative environment. Get started: nvda.ws/3cTGMbx
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Today Sunday at 10:00AM EST we have a full day tutorial about #selfdriving at #cvpr2021! Overview on the cutting edge of self-driving (perception prediction, planning, maps, simulation, etc). Come learn how you can change the world with self-driving!
cvpr2021.waabi.ai -
Sanja Fidler reposted
Arash Vahdat
@ArashVahdat
📢 Latent Score-based Generative Model (LSGM)
Score-based generative models (SGMs) are applied directly in data space & often require 1000s of network evaluations for sampling. We introduce LSGM for training SGMs in a latent space.
w/ @karsten_kreis(eq. contrib.) & @jankautz -
Sanja Fidler reposted
Waabi
@Waabi_ai
Tune in to @TechCrunch Sessions: Mobility tomorrow (6:05pm ET) to watch our founder @RaquelUrtasun speak about launching Waabi and how AI can speed up the commercialization of AVs. Register here:
techcrunch.com/events/tc-sessions-mobility-2021/ -
Sanja Fidler reposted
Waabi
@Waabi_ai
Hi! We’re Waabi - a new company founded by @raquelurtasun that is building the next generation of self-driving technology. Learn more or join us at waabi.ai.
#ai #selfdriving #innovation
www.globenewswire.com/news-release/2021/06/08/2243365/0/e… -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Super excited to share the next step of my #ai journey with the founding of my own company and an incredible $83.5M Series A financing round. Join me @waabi_ai and help us build the next generation of self-driving tech, check out waabi.ai
www.globenewswire.com/news-release/2021/06/08/2243365/0/e… -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
#cvpr21 is around the corner - @NVIDIA researchers have nearly 30 accepted papers that range from simulating dynamic driving environments, to powering neural architecture search for medical imaging. Read more: nvda.ws/3g89dTX
-
Sanja Fidler reposted
Tianmin Shu
@tianminshu
We will host an ICRA workshop on Social Intelligence in Humans and Robots on June 4th (Fri) at 9:30 am ET, where we will bring together experts from robotics, AI, and cognitive science to discuss the advances and challenges on engineering socially intelligent robots and machines
-
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
🚨The source code for our #cvpr2021 Oral paper, Neural Geometric Level of Detail is now available on GitHub under the MIT License!
GitHub: github.com/nv-tlabs/nglod
Project Web: nv-tlabs.github.io/nglod/
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢 Our new @NVIDIAAI paper is up on arXiv!
I'm happy to share Neural Geometric Level of Detail, which enables the first ⚡real-time rendering of high quality neural #3d SDFs with a sparse octree feature volume and a tiny MLP!
nv-tlabs.github.io/nglod/ -
Sanja Fidler
@FidlerSanja
Code for Neural LOD is released!
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
🚨The source code for our #cvpr2021 Oral paper, Neural Geometric Level of Detail is now available on GitHub under the MIT License!
GitHub: github.com/nv-tlabs/nglod
Project Web: nv-tlabs.github.io/nglod/
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢 Our new @NVIDIAAI paper is up on arXiv!
I'm happy to share Neural Geometric Level of Detail, which enables the first ⚡real-time rendering of high quality neural #3d SDFs with a sparse octree feature volume and a tiny MLP!
nv-tlabs.github.io/nglod/ -
Sanja Fidler reposted
Andrew Liao
@andrewliao11
We are pleased to share our upcoming CVPR 2021 Oral paper “Towards Good Practices for Efficiently Annotating Large-Scale Image Classification Datasets” 🔥
What matters when annotating a Large-Scale Image Classification Dataset? We provide you the cookbook of it! 🧑🍳👩🍳👇 w -
Sanja Fidler reposted
University of Toronto
@UofT
Once @UofTCompSci students at #uoft, Ivan Zhang, Aidan Gomez & Nick Frosst are now co-founders of an #ai startup that has big-name backers, including top researchers at Google, Apple, Uber & Nvidia. 💻 uoft.me/6Hp
-
Sanja Fidler reposted
Yuxuan Zhang
@YuxuanZ84175540
Drop by our oral session @iclr_conf and see how we can reconstruct 3D models from real images and control StyleGAN via interpretable 3D properties!
Paper: arxiv.org/abs/2010.09125
ICLR Oral Session 5: 4th May 3:15 p.m. - 3:30 p.m. EDT
Yuxuan Zhang
@YuxuanZ84175540
Excited to share our @nvidiaai work on reconstructing high-quality 3D models from monocular photographs and making StyleGAN a controllable neural renderer.
@ChenWenzheng @HuanLing6 @JunGao33210520 Yinan Zhang, Antonio Torralba @FidlerSanja
paper: arxiv.org/abs/2010.09125 -
Sanja Fidler
@FidlerSanja
Who is best to deliver advanced language models to every business than the co-inventor of Transformers, @AidanNGomez ! Super excited to follow your success @CohereAI, @AidanNGomez, @nickfrosst (co-inventor of Capsules) and @1vnzh !
Jordan Jacobs
@JordanJacobs10
Today @CohereAI comes out of stealth to bring the most advanced AI language models to every business via a simple API. Founded by transformer co-creator @AidanNGomez, @geoffreyhinton protege @nickfrosst & @1vnzh. www.theglobeandmail.com/business/article-toronto-startup-…
-
Sanja Fidler reposted
Jordan Jacobs
@JordanJacobs10
Today @CohereAI comes out of stealth to bring the most advanced AI language models to every business via a simple API. Founded by transformer co-creator @AidanNGomez, @geoffreyhinton protege @nickfrosst & @1vnzh. www.theglobeandmail.com/business/article-toronto-startup-…
-
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
Excited to share our work DriveGAN, accepted at #cvpr2021 as an oral! We introduce an interactive neural simulator trained on real videos. Joint work with @PhilionJonah @abtorralba @FidlerSanja @NVIDIAAI
Webpage: nv-tlabs.github.io/DriveGAN/
Paper: arxiv.org/abs/2104.15060 -
Sanja Fidler
@FidlerSanja
Releasing DriveGAN, a controllable neural simulator -- simulating driving/navigation environments directly in pixel space, enabling ego agents to drive and re-drive through scenes for which we can control weather, objects, etc. @seungkim0123 @PhilionJonah @abtorralba @NVIDIAAI
Seung Wook Kim
@seungkim0123
Excited to share our work DriveGAN, accepted at #cvpr2021 as an oral! We introduce an interactive neural simulator trained on real videos. Joint work with @PhilionJonah @abtorralba @FidlerSanja @NVIDIAAI
Webpage: nv-tlabs.github.io/DriveGAN/
Paper: arxiv.org/abs/2104.15060 -
Sanja Fidler reposted
Michael Zhang
@mzhangio
*First tweet* 🍳
Excited to share our work on Personalized Federated Learning with First Order Model Optimization (FedFomo)! (also at ICLR 2021)
arxiv.org/abs/2012.08565
Impossible without @currenator, @FidlerSanja, @syeung10, @ALVAREZ_JOSEM
More motivation + method below: -
Sanja Fidler reposted
Zhiding Yu
@ZhidingYu
Wondering how to improve detection under long-tailed distributions? Our recent work shows that simple object-level resampling can lead to significantly better results.
arXiv: arxiv.org/abs/2104.05702
@nadinechang430 Yu-Xiong Wang Anima Anandkumar @FidlerSanja @ALVAREZ_JOSEM
(1/4) -
Sanja Fidler reposted
Masha Shugrina
@_shumash
We just released a major update to the #nvidiakaolin library and a new companion #nvidiaomniverse app for 3D #deeplearning. We are actively adding features, and our in-browser remote 3D checkpoint visualizer was just pushed to github today. Start using this and send us feedback!
Sanja Fidler
@FidlerSanja
#nvidiakaolin library for 3D Deep Learning, which has recently been updated, now comes with an @nvidiaomniverse Kaolin App for super cool visualization features! Most of the 3D visualizations in our recent papers are created with it. Check it out!
NVIDIA Omniverse
@nvidiaomniverse
Researchers can now take 3D #deeplearning into #nvidiaomniverse with the #nvidiakaolin app - advancing 3D training #visualization & data generation.
Combined with the updated #kaolinlibrary, research accelerates for a wide range of 3D applications.
🔗: nvda.ws/3mYvTcj -
Sanja Fidler reposted
Darrin Johnson
@darrinpjohnson
Researchers can now take 3D #deeplearning into #nvidiaomniverse with the #nvidiakaolin app - advancing 3D training #visualization & data generation. Combined with the updated #kaolinlibrary, research accelerates for a wide range of 3D applications. share.nvidia.com/Article/Redirect/e48ce223-31dc-4a5b-9ee0…
-
Sanja Fidler
@FidlerSanja
#nvidiakaolin library for 3D Deep Learning, which has recently been updated, now comes with an @nvidiaomniverse Kaolin App for super cool visualization features! Most of the 3D visualizations in our recent papers are created with it. Check it out!
NVIDIA Omniverse
@nvidiaomniverse
Researchers can now take 3D #deeplearning into #nvidiaomniverse with the #nvidiakaolin app - advancing 3D training #visualization & data generation.
Combined with the updated #kaolinlibrary, research accelerates for a wide range of 3D applications.
🔗: nvda.ws/3mYvTcj -
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
Researchers can now take 3D #deeplearning into #nvidiaomniverse with the #nvidiakaolin app - advancing 3D training #visualization & data generation.
Combined with the updated #kaolinlibrary, research accelerates for a wide range of 3D applications.
🔗: nvda.ws/3mYvTcj -
Sanja Fidler reposted
Fuji Tsang Clément
@Caenorst
Excited to announce the release of the Omniverse Kaolin App, with an update of #kaolinlibrary with new features and speedups!
NVIDIA Omniverse
@nvidiaomniverse
Researchers can now take 3D #deeplearning into #nvidiaomniverse with the #nvidiakaolin app - advancing 3D training #visualization & data generation.
Combined with the updated #kaolinlibrary, research accelerates for a wide range of 3D applications.
🔗: nvda.ws/3mYvTcj -
Sanja Fidler reposted
AK
@_akhaliq
Semantic Segmentation with Generative Models: Semi-Supervised Learning and Strong Out-of-Domain Generalization
pdf: arxiv.org/pdf/2104.05833.pdf
abs: arxiv.org/abs/2104.05833
project page: nv-tlabs.github.io/semanticGAN/ -
Sanja Fidler reposted
AK
@_akhaliq
DatasetGAN: Efficient Labeled Data Factory with Minimal Human Effort
pdf: arxiv.org/pdf/2104.06490.pdf
abs: arxiv.org/abs/2104.06490 -
Sanja Fidler
@FidlerSanja
We show GANs to be excellent representation learners! With only 16 images labeled with pixel-wise information, we can beat fully supervised method that require 100x more labeled data! datasetGAN takes a few manually labeled examples and synthesizes large labeled datasets.
Sanja Fidler
@FidlerSanja
Because we now need very few labels, we go and label each example to a super high level of detail. Below we show cars manually segmented with object part labels. The picture shows our ENTIRE labeled dataset.
Sanja Fidler
@FidlerSanja
We showcase the ability of datasetGAN to label object parts at a very high detail by combining it with our ICLR paper (arxiv.org/abs/2010.09125) to be able to predict animatable 3D models directly from images; which we call GANverse3D.
blogs.nvidia.com/blog/2021/04/16/gan-research-knight-ride…
@NVIDIAAI -
Sanja Fidler
@FidlerSanja
Using GANs for semantic segmentation! semanticGAN exploits very little labeled data to achieve good performance and shows super strong generalization to very different domains, eg by training on real faces & testing on cartoon faces, or going from CT scans of organs to MRI scans.
Daiqing Li
@lidaiqing
Exciting to share our work SemanticGAN which exploits generative models for pixel-wise discriminative tasks with strong generalization ability and high label efficiency. Joint work with Junlin Yang, Karsten Kreis, Antonio Torralba, @FidlerSanja , page: nv-tlabs.github.io/semanticGAN/
-
Sanja Fidler
@FidlerSanja
Check out our work on GANverse3D!
NVIDIA AI
@NVIDIAAI
NVIDIA Researchers transform the Knight Rider K.I.T.T. car from a 2D-to-3D model with GANverse3D. Learn more in our blog: nvda.ws/3uTiuFs. @nvidiaomniverse
-
Sanja Fidler reposted
Andreas Schilling 🇺🇦
@aschilling
.@FidlerSanja and her team @nvidiaomniverse / @UofTCompSci did publish a research paper showing #ganverse3d, a controllable 3D “neural renderer” that is currently trained on cars – 2D picture to 3D modell in a few milliseconds.
#gtc21
www.hardwareluxx.de/index.php/news/software/anwendungprog… -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Join me now for an exciting talk at #gtc21 on the future of #selfdriving. I'll be revealing our latest and greatest on how to build a super high fidelity close loop simulator. I will be answering live 1:1 questions during my talk. See you there!
-
Sanja Fidler reposted
Jun Gao
@JunGao33210520
Excited to see our work was featured in Jensen's GTC keynote. @NVIDIAAI @nvidiaomniverse @nvidia
Joint work with @YuxuanZ84175540 @ChenWenzheng @HuanLing6 @FidlerSanja -
Sanja Fidler reposted
Krishna Murthy
@_krishna_murthy
At @iclr_conf, we're excited to present 𝗴𝗿𝗮𝗱𝗦𝗶𝗺, a differentiable simulator 👇
gradSim enables visuomotor control and physical parameter estimation directly from images/videos
arxiv.org/abs/2104.02646
Co-led with @milesmacklin
gradsim.github.io -
Sanja Fidler reposted
Vector Institute
@VectorInst
Register today for #gtc21, featuring Vector co-founders @GeoffreyHinton, @RaquelUrtasun, and @FidlerSanja; 1200+ sessions on #appliedai; and more. April 12-16; registration is free.
NVIDIA GTC
@NVIDIAGTC
Registration is now open for #gtc21! Get access to over 1200 free sessions, grow your expertise, and help shape what’s next with fellow technologists, #ai innovators, and #creators April 12-16. Register now. nvda.ws/3sR0kDh
-
Sanja Fidler
@FidlerSanja
NVIDIA Toronto is looking for engineers to work on 3D deep learning applications. @gavrielstate is an excellent manager and we collaborate closely on many projects. Please apply!
Gavriel State
@gavrielstate
My team at #nvidia is hiring!
Get involved in #nvidiaomniverse efforts around #syntheticdata, collaborate with our Toronto AI research team around the Kaolin 3D #deeplearning library and more!
Apply here nvidia.wd5.myworkdayjobs.com/NVIDIAExternalCareerSite/job… to work at the intersection of graphics and AI! -
Sanja Fidler reposted
Despoina Paschalidou
@paschalidoud_1
📢Our new #cvpr2021 paper with @angeloskath, @FidlerSanja and Andreas Geiger is on arXiv!
We present Neural Parts, a novel 3D primitive representation that is not limited to a specific family of shapes but instead yields arbitrarily complex primitives.
paschalidoud.github.io/neural_parts -
Sanja Fidler
@FidlerSanja
Check out Neural Parts -- unsupervised part discovery and reconstruction from images! Great collaboration with @paschalidoud_1 @angeloskath and Andreas Geiger
Despoina Paschalidou
@paschalidoud_1
📢Our new #cvpr2021 paper with @angeloskath, @FidlerSanja and Andreas Geiger is on arXiv!
We present Neural Parts, a novel 3D primitive representation that is not limited to a specific family of shapes but instead yields arbitrarily complex primitives.
paschalidoud.github.io/neural_parts -
Sanja Fidler
@FidlerSanja
Congrats to @amlankar95 and @SirrahChan on receiving the RBC fellowship!
Vector Institute
@VectorInst
Congratulations to the 2021 RBC Post-Doctoral & Graduate Fellowship winners, including Vector's Harris Chan & Amlan Kar. These Comp Sci PhD students work with Vector faculty members Jimmy Ba & @FidlerSanja respectively. We look forward to all that you will achieve! #ai #research
-
Sanja Fidler reposted
NVIDIA Omniverse
@nvidiaomniverse
#gtc21 registration is now open! From April 12-16 you will have access to over 1200 free sessions as well as the opportunity to grow your expertise & help shape what’s next with fellow technologists, #ai innovators, & #creators. Register today: nvda.ws/3l0GUcp
-
Sanja Fidler reposted
dara khosrowshahi
@dkhos
Thanks for your thought leadership, pushing the team to excel and always being UberFirst, @RaquelUrtasun. Has been an amazing ride. So excited for all the incredible things you'll continue to accomplish!
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Today is my last day @Uber after 4 wonderful years. Thank you @dkhos for your support and leadership. Solving #selfdriving is my life's passion and I'm super excited for what comes next. Will share an update soon. Stay tuned!
-
Sanja Fidler reposted
James Tompkin
@jtompkin
AI for Content Creation Workshop is back at #cvpr2021. visual.cs.brown.edu/aicc2021 4/8-page paper submissions open, due March 8th; best paper prize!
Tentative speakers: @talidekel cs.utexas.edu/users/grauman/ @MattNiessner @deviparikh @SergeyTulyakov @RaquelUrtasun @kemelmi @jon_barron ! -
Sanja Fidler reposted
Harrison Kinsley
@Sentdex
Using code adapted from GameGAN (github.com/nv-tlabs/GameGAN_code), @daniel_kukiela and I have been tinkering. GameGAN is where the neural network *is the environment* that you play within. The NN makes the rules and visuals. Here's a quick example of us playing a racecar game we trained:
-
Sanja Fidler reposted
University of Toronto
@UofT
Why do we need #womeninscience? Leading #uoftwomeninstem have a special message to inspire the next generation of scientists on International Day of Women & Girls in Science.
This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
Here's a better version of this same gif but with numerical speed differences :)
-
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
By popular demand, here are some motivations and insights that enable significant speedups for Neural LOD:
-
Sanja Fidler reposted
morgan
@CasualEffects
New work with @yongyuanxi, @JoeyLitalien, @FidlerSanja, @DerekRenderling, @_AlecJacobson, Kangxue Yin, Karsten Kreis, & Charles Loop:
Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes, 2021
arxiv.org/abs/2101.10994 -
Sanja Fidler reposted
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢 Our new @NVIDIAAI paper is up on arXiv!
I'm happy to share Neural Geometric Level of Detail, which enables the first ⚡real-time rendering of high quality neural #3d SDFs with a sparse octree feature volume and a tiny MLP!
nv-tlabs.github.io/nglod/ -
Sanja Fidler
@FidlerSanja
Real-time rendering of neural implict representations! 2-3 orders of magnitude faster than previous work. @NVIDIAAI #3d
Towaki Takikawa / 瀧川永遠希
@yongyuanxi
📢 Our new @NVIDIAAI paper is up on arXiv!
I'm happy to share Neural Geometric Level of Detail, which enables the first ⚡real-time rendering of high quality neural #3d SDFs with a sparse octree feature volume and a tiny MLP!
nv-tlabs.github.io/nglod/ -
Sanja Fidler reposted
Seung Wook Kim
@seungkim0123
Code for "Learning to Simulate Dynamic Environments with GameGAN"
(CVPR 2020, nv-tlabs.github.io/gameGAN/)
is now available!
github.com/nv-tlabs/GameGAN_code
@FidlerSanja @PhilionJonah @zhouyh7 @abtorralba -
Sanja Fidler
@FidlerSanja
Code for GameGAN is now released! @NVIDIAAI @seungkim0123 @PhilionJonah @zhouyh7 @abtorralba #artificialintelligence #gaming
project page: nv-tlabs.github.io/gameGAN/
code: github.com/nv-tlabs/GameGAN_code
video: youtu.be/4OzJUNsPx60
Seung Wook Kim
@seungkim0123
Code for "Learning to Simulate Dynamic Environments with GameGAN"
(CVPR 2020, nv-tlabs.github.io/gameGAN/)
is now available!
github.com/nv-tlabs/GameGAN_code
@FidlerSanja @PhilionJonah @zhouyh7 @abtorralba -
Sanja Fidler reposted
Yuxuan Zhang
@YuxuanZ84175540
Excited to share that our paper gets accepted to ICLR as an oral presentation!
Yuxuan Zhang
@YuxuanZ84175540
Excited to share our @nvidiaai work on reconstructing high-quality 3D models from monocular photographs and making StyleGAN a controllable neural renderer.
@ChenWenzheng @HuanLing6 @JunGao33210520 Yinan Zhang, Antonio Torralba @FidlerSanja
paper: arxiv.org/abs/2010.09125 -
Sanja Fidler reposted
Maarten van Smeden
@MaartenvSmeden
How to become a SUCCESSFUL academic: a guide 1/n
-
Sanja Fidler reposted
Avik Pal
@avikpal1410
👇 got accepted at #iclr2021!!
Avik Pal
@avikpal1410
Excited to share our @UofT & @VectorInst work "Emergent Road Rules in Multi-Agent Driving Environments" w/ @PhilionJonah, Andrew Liao & @FidlerSanja.
Paper: arxiv.org/abs/2011.10753
Project Page: fidler-lab.github.io/social-driving/
Code: github.com/fidler-lab/social-driving
[1/5] -
Sanja Fidler
@FidlerSanja
University of Toronto CS has open faculty positions in computer vision as well as machine learning (at different levels of seniority). Please apply, and join us at UofT! @UofTCompSci @UofT #computervision #machinevision #machinelearning
Apply here: web.cs.toronto.edu/employment-opportunities -
Sanja Fidler reposted
U of T Department of Computer Science
@UofTCompSci
Join us at #uoft!
We are currently accepting applications for multiple tenure-stream #facultyjobs starting July 1, 2021.
Learn more ➡️ uoft.me/6hV -
Sanja Fidler reposted
Nando de Freitas
@NandoDF
Can AI researchers please tweet: I am against racism, sexism, bullying and cancelling, and I believe in improving diversity, equity and inclusion in our AI community. We need to hear your voices! The students and the public need to know what most believe.
-
Sanja Fidler
@FidlerSanja
Congrats Rev and team on creating this amazing platform! Super excited to work with you!
-
Sanja Fidler
@FidlerSanja
Open beta of NVIDIA's Omniverse is now available!
Many of 3D visualizations and demos in our recent papers are done in Omniverse. Super awesome.
NVIDIA Workstation
@NVIDIAworkstatn
Explore a new world of virtual collaboration and photorealistic simulation with the #nvidiaomniverse open beta, available now. Learn more about the open beta release: nvda.ws/3mkEBQl
-
Sanja Fidler reposted
Dima Damen
@dimadamen
To be disgusted by this thread is the least one can do.
To be surprised means you have been lucky enough not to face this type of blatant sexism and discrimination.
To be fearful for future generations is understandable.
To be only sad is a luxury we cannot afford.
#iamadoctor -
Sanja Fidler reposted
Huan Ling
@HuanLing6
Excited to share our @NVIDIAAI paper at #neurips2020.
Variational Amodal Object Completion
w/ @davidjesusacu Karsten Kreis @seungkim0123 @FidlerSanja
Poster : Dec 11th, 2020 @ 00:00 – 02:00 EST
at poster session 7 ( Town D1 - Spot A1 )
Project page: www.cs.toronto.edu/~linghuan/vae_mask/ -
Sanja Fidler
@FidlerSanja
Check out our work on amodal instance segmentation with application to image editing, at NeurIPS! @NVIDIAAI #neurips2020
-
Sanja Fidler
@FidlerSanja
Check out our DefTet work at NeurIPS 😊
NVIDIA AI
@NVIDIAAI
Check out our researchers' incredible work at #neurips2020! @JunGao33210520 @ChenWenzheng @TommyX058 @_AlecJacobson @CasualEffects @FidlerSanja
-
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
Interested in joining NVIDIA's Research team? Join us tomorrow from 12-2 PM PST to hear NVIDIA leaders share how you can make a direct impact in solving some of the world's hardest problems. #neurips2020 @FidlerSanja @jankautz @ctnzr @AnimaAnandkumar @ALVAREZ_JOSEM
-
Sanja Fidler
@FidlerSanja
Congratulations Krishna!!! It's been great collaborating with you and excited to work together in the future! @NVIDIAAI
Krishna Murthy
@_krishna_murthy
Honored and thrilled to have received an @NVIDIAAI PhD fellowship!! blogs.nvidia.com/blog/2020/12/04/graduate-fellowships-gpu…
I'd like to acknowledge the support of my advisor @duckietown_coo and close mentors @DerekRenderling @FidlerSanja who make work so much fun :) @UMontrealDIRO @Mila_Quebec @MontrealRobots -
Sanja Fidler
@FidlerSanja
This is super cool
-
Sanja Fidler reposted
Alexandre Moufarek
@amoufarek
Heads up urban open world #gamedevs!
When AI agents are trained to drive as fast as possible from A to B without colliding: road rules emerge! (@avikpal1410, @PhilionJonah, @andrewliao11 & @FidlerSanja, 2020)
Project: fidler-lab.github.io/social-driving/
Paper: arxiv.org/abs/2011.10753 -
Sanja Fidler
@FidlerSanja
Check out our @NVIDIAAI work on learning a neural controller for physics-based character motion! Unicon models thousands of skills with a single network, transfers to unseen motions, characters and perturbation types, and supports character #animation from videos.
tingwu.wang
@TingwuWang
Excited to share
"UniCon: Universal Neural Controller For Physics-based Character Motion"
@nvidia @VectorInst @UofT
nv-tlabs.github.io/unicon/
By systematically reviewing the RL adaptation, we greatly close the gap between physics-based animation and real-life applications (1/3) -
Sanja Fidler
@FidlerSanja
We have an exciting group of ML/Vision/robotics/graphics faculty @UofTCompSci. For interested #students, please apply by Dec 1 on web.cs.toronto.edu/graduate/admissions and join us for grad school! 😊 #machinelearning #artificialintelligence #deeplearning
Animesh Garg
@animesh_garg
we have a great group of AI/ML in {robotics + vision}
consider applying and do mention the specific names in your application before Dec 1
@FidlerSanja @RaquelUrtasun @florian_shkurti @igilitschenski and myself
Sanja Fidler
@FidlerSanja
Many of the faculty and thus their students are affiliated with @VectorInst, having access to impressive compute for deep learning. Apart from a chilly winter, Toronto is an awesome city to live in!
-
Sanja Fidler reposted
CIFAR
@CIFAR_News
Canada is a destination for talent in artificial intelligence & innovation thanks to our vibrant AI ecosystem: Elissa Strome, ED of the CIFAR Pan-Canadian AI Strategy on the Strategy's impact to date at @Mila_Quebec @VectorInst @AmiiThinks
www.cifar.ca/cifarnews/2020/11/17/a-look-toward-2030 -
Sanja Fidler
@FidlerSanja
Really excited about our new driving environments for RL! By only training the agents to drive as fast as possible from A to B without colliding, road rules like following lanes,traffic signs and right of way emerge! Code released! @UofT @UofTCompSci @researchuoft @VectorInst
Avik Pal
@avikpal1410
Excited to share our @UofT & @VectorInst work "Emergent Road Rules in Multi-Agent Driving Environments" w/ @PhilionJonah, Andrew Liao & @FidlerSanja.
Paper: arxiv.org/abs/2011.10753
Project Page: fidler-lab.github.io/social-driving/
Code: github.com/fidler-lab/social-driving
[1/5] -
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
NVIDIA researchers continue to push the boundaries of technology in machine learning, self-driving cars, robotics, graphics, and more. Our researchers will present 28 papers at #neurips2020 from 12/6 to 12/12. Click here to see our lineup of papers: nvda.ws/3fzYfGm
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Want to reduce the latency of your self-driving autonomy system by 100ms? Check our work to learn how to perform perception without waiting for the LiDAR sweep to be built!
Davi Frossard
@davi_frossard
Existing LiDAR perception systems wait 100ms just to build a sweep. In our #corl2020 paper StrObe we instead do streaming detection from LiDAR packets and achieve an end-to-end latency of 21ms!
Paper: arxiv.org/abs/2011.06425
Video: youtu.be/Gy97AnWdGn8
Session: Today@14:50 EST -
Sanja Fidler reposted
arXiv Daily
@Arxiv_Daily
Learning Deformable Tetrahedral Meshes for 3D Reconstruction
deepai.org/publication/learning-deformable-tetrahedral-me…
by @gaojun4ever et al. including @mcgenergy, @FidlerSanja
#lossfunction #machinelearning -
Sanja Fidler
@FidlerSanja
Our #neurips2020 work "DefTet" is a new #3d DL representation! We predict tetrahedral meshes directly from input (point cloud, single or multitview images) -- accurate & fast! @NVIDIAAI #machinelearning
Project page: nv-tlabs.github.io/DefTet/
code coming soon
Jun Gao
@JunGao33210520
Excited to share our @NVIDIAAI work DefTet, which makes tetrahedral meshes, one of the major 3D representations in graphics and physics-based simulation, amenable to Deep Learning.
@ChenWenzheng @TommyX058 @_AlecJacobson @CasualEffects @FidlerSanja
Paper: arxiv.org/abs/2011.01437 -
Sanja Fidler reposted
NVIDIA Workstation
@NVIDIAworkstatn
Built on #nvidiaomniverse and running on a single GeForce RTX 3090, “Marbles at Night” showcases how creative professionals can collaborate using true-to-life simulations with photorealistic lighting, accurate reflections, and real-time ray tracing. nvda.ws/2TOoDSM
-
Sanja Fidler reposted
ViktorM🇺🇦
@viktor_m81
Do you want to train tasks on a single PC that was possible to solve only using big CPU clusters?
Proud to share the project I've been working on for the last 2 years, end-to-end GPU accelerated simulator for RL and control: developer.nvidia.com/isaac-gym -
Sanja Fidler reposted
Blog: Top 10 Tech, Innovation and Startup Tweets
@Top10TechTweets
October 2020: Top 10 #technology And #innovation Tweets
www.joshinishad.blog/2020/10/october-2020-vol-18-10-impor…
Content credit/thanks:
@CoinDesk
@anandmahindra
@CiscoUmbrella
@MIT_CSAIL
@svlevine
@QuickTake
@TechParida
@BBCWorld
@BozzatoFabrizio
@Paula_Piccard
@Cointelegraph
@FidlerSanja
@EricNewcomer
@bitcoinnews -
Sanja Fidler
@FidlerSanja
Our PKL metric to evaluate object detectors is now available on the nuScenes leaderboard! @NVIDIAAI #selfdriving
Jonah Philion
@PhilionJonah
Our metric "Planning KL-Divergence" is now on the @motionaldrive nuScenes leaderboard! Check it www.nuscenes.org/object-detection
IROS 2020 Workshop paper w Yiluan Holger @beijbom @FidlerSanja arxiv.org/abs/2010.09350
@CVPRConf 2020 paper w @amlankar95 @FidlerSanja arxiv.org/abs/2004.08745 -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Sharing really exciting work on Product quantization for neural network compression. Code available. Check it out!
-
Sanja Fidler
@FidlerSanja
Super excited about our @NVIDIAAI work on single-image #3d reconstruction that utilizes #stylegan as a multi-view training data generator. Only a few minutes of labeling required for results below! We also turn StyleGAN into a controllable 3D neural renderer. #machinelearning
Yuxuan Zhang
@YuxuanZ84175540
Excited to share our @nvidiaai work on reconstructing high-quality 3D models from monocular photographs and making StyleGAN a controllable neural renderer.
@ChenWenzheng @HuanLing6 @JunGao33210520 Yinan Zhang, Antonio Torralba @FidlerSanja
paper: arxiv.org/abs/2010.09125
Sanja Fidler
@FidlerSanja
Special mention to Alex Zhang @YuxuanZ84175540, who did this work during his @NVIDIAAI internship. Alex is an undergraduate student from @UWaterloo -- applying for graduate school this year. Awesome job, Alex!!
-
Sanja Fidler reposted
-
Sanja Fidler reposted
CoCoSci MIT
@MITCoCoSci
We also build a multi-agent virtual platform, where you can train and test your favorite agents with other agents or with real humans in realistic household environments. arxiv: arxiv.org/abs/2010.09890 code: github.com/xavierpuigf/watch_and_help video: youtu.be/lrB4K2i8xPI
-
Sanja Fidler
@FidlerSanja
Really exciting multi-agent collaboration environment!
@xavierpuigf, @tianminshu, @ShuangL13799063, Zilin Wang, Josh Tenenbaum, Antonio Torralba #machinelearning
Xavier Puig
@xavierpuigf
Excited to share “Watch-And-Help: A Challenge for Social Perception and Human-AI Collaboration”!
We propose a challenge where agents need to infer human goals in a household and help perform them efficiently.
arxiv: arxiv.org/abs/2010.09890
code: github.com/xavierpuigf/watch_and_help -
Sanja Fidler reposted
Dima Damen
@dimadamen
Just released! code and trained models for #epickitchens-100 baselines
action recognition [TSN/TSM/TRN/TBN/Slowfast -Credit @will_price_94 @e_kazakos]
github.com/epic-kitchens/C1-Action-Recognition
and action anticipation [RULSTM -credit @anfurnari]
github.com/epic-kitchens/C3-Action-Anticipation
Dima Damen
@dimadamen
“Rescaling Egocentric Vision” describing EPIC-KITCHENS-100, collection and challenges is now on ArXiv: arxiv.org/abs/2006.13256
New annotation pipeline builds on ‘Pause-and-Talk’ narration interface, resulting in increased density and quality of annotations, and 6 challenges. -
Sanja Fidler
@FidlerSanja
Code for Lift Splat paper is released!
Jonah Philion
@PhilionJonah
Code for training a lift-splat-shoot model for camera-only BEV vehicle segmentation is released today! Link below
code: github.com/nv-tlabs/lift-splat-shoot
site: nv-tlabs.github.io/lift-splat-shoot/
paper: arxiv.org/abs/2008.05711
@eccvconf @NVIDIAAI @UofTCompSci @VectorInst -
Sanja Fidler reposted
Ronnie Clark
@ronnieclark__
Great (anonymous) idea from #iclr2021: use a GAN to synthesize multi-view images then feed them to an inverse graphics network to recover shape, texture and lighting. This can in turn be used to disentangle the latent code etc. openreview.net/pdf?id=yWkP7JuHX1 #machinelearning
-
Sanja Fidler reposted
-
Sanja Fidler
@FidlerSanja
Excited to present our work on Federated Simulation @MICCAI2020! We model an organ in 3D, and use a physics-driven simulator to synthesize CT volumes along with labels, mimicking the target imaging device. @NVIDIAAI
paper: arxiv.org/abs/2009.00668
webpage: nv-tlabs.github.io/fed-sim/
Daiqing Li
@lidaiqing
Our work "Federated Simulation for Medical Imaging" @amlankar95 nishant @affrangi @FidlerSanja will be presented @MICCAI2020 😀
oral session: 11:00 AM - 11:30 AM EDT (Tue, 6, Oct)
poster session: 12:30 PM - 2:00 PM (Tue, 6, Oct).
Page: nv-tlabs.github.io/fed-sim/ @NVIDIAAI -
Sanja Fidler reposted
Future of Film
@FutureofFilm_
The GPU Technology Conference from @nvidia has kicked off! Who else is tuning in over the next couple of days?
Really looking forward to seeing chats with @HaZ_Dulull and @FidlerSanja in particular, two innovators constantly pushing boundaries!
#gtc20
NVIDIA
@nvidia
Watch our #gtc20 Keynote to find out the latest breakthroughs for #ai, deep learning, and our company this year and beyond: nvda.ws/30yoFSz
-
Sanja Fidler reposted
University of Toronto
@UofT
Later this afternoon, the University of Toronto will make a historic announcement that will have transformative impact – join us and learn more. www.utoronto.ca/news/university-toronto-receives-single-l…
-
Sanja Fidler
@FidlerSanja
Code for the Deformable Grid (DefGrid) paper is now released -- instead of processing images on fixed grids as in traditional pipelines, we learn to deform the grid in end2end manner. Plug&play module, nice accuracy boosts!
Code: github.com/fidler-lab/defgrid-release
Jun Gao
@JunGao33210520
Excited to release the code of our @eccvconf paper (DefGrid)! @FidlerSanja @zianwang97
We provide three examples, and we hope it can benefits other Computer Vision tasks as well!
Code: github.com/fidler-lab/defgrid-release/tree/master
Paper: arxiv.org/abs/2008.09269
Project: www.cs.toronto.edu/~jungao/def-grid/ -
Sanja Fidler reposted
Jun Gao
@JunGao33210520
Excited to release the code of our @eccvconf paper (DefGrid)! @FidlerSanja @zianwang97
We provide three examples, and we hope it can benefits other Computer Vision tasks as well!
Code: github.com/fidler-lab/defgrid-release/tree/master
Paper: arxiv.org/abs/2008.09269
Project: www.cs.toronto.edu/~jungao/def-grid/ -
Sanja Fidler reposted
Bryan Catanzaro
@ctnzr
Jensen Huang is a time machine who lives in the future. So much of @nvidia’s great success over the past 5 years happened because of the strategic decisions that he made 15 years ago along with his continuing investments to bring them from vision to reality.
Stratechery
@stratechery
Nvidia's Integration Dreams
Nvidia's acquisition of ARM only makes sense from a financial perspective, unless you buy Jensen Huang's datacenter dreams.
stratechery.com/2020/nvidias-integration-dreams/ -
Sanja Fidler
@FidlerSanja
So exciting to be NVIDIAn in this historic moment!!
NVIDIA
@nvidia
NVIDIA to Acquire Arm for $40 Billion, Creating World’s Premier Computing Company for the Age of AI. nvda.ws/2Rk5RRY
-
Sanja Fidler reposted
Patrick Moorhead
@PatrickMoorhead
Talked with both @NVIDIA's Jensen Huang and @ARM's @simonsegars about the $40B deal. www.forbes.com/sites/patrickmoorhead/2020/09/13/its-offic…
-
Sanja Fidler reposted
U of T Department of Computer Science
@UofTCompSci
With a tool to rapidly and accurately label large amounts of data, Associate Professor @FidlerSanja is helping speed up researchers' AI innovations. www.cifar.ca/cifarnews/2020/09/09/sanja-fidler-is-taking-… via @CIFAR_News
-
Sanja Fidler reposted
NVIDIA
@nvidia
Registration for #gtc20 this fall is now open! Join us online on Oct. 5-9 for global insights and innovations. Early-bird pricing ends Sept. 25. Register now: nvda.ws/2EbfYVX
-
Sanja Fidler reposted
NVIDIA GeForce
@NVIDIAGeForce
8K gaming has arrived!
GeForce GTX 3090 is the first graphics card that can play, capture and watch in 8K.
That’s 4X as many pixels as 4K, and 16X as many pixels as 1080p, for incredible levels of detail.
www.nvidia.com/en-us/geforce/news/geforce-rtx-3090-8k-hdr…This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
Jeremy Howard
@jeremyphoward
I imagine pretty much everyone doing deep learning that can afford an RTX 3090 is going to be buying at least one. It's much better than the Titan RTX, and 30% cheaper!
www.anandtech.com/show/16057/nvidia-announces-the-geforce… -
Sanja Fidler reposted
NVIDIA GeForce
@NVIDIAGeForce
New RT Cores.
New Tensor Cores.
Superfast GDDR6X memory.
The all-new GeForce RTX 30 Series is The #ultimateplay.
www.nvidia.com/en-us/geforce/news/introducing-rtx-30-seri…This media couldn't be displayed. See it on the original post. -
Sanja Fidler reposted
ML and Data Projects To Know
@PTKDaily
📙 Beyond Fixed Grid: Learning Geometric Image Representation with a Deformable Grid
Authors: @JunGao33210520 @zianwang97 Jinchen Xuan, @FidlerSanja
Featured in PTK #49 mailchi.mp/amplifypartners/ptk47-12333350
Paper: arxiv.org/pdf/2008.09269.pdf -
Sanja Fidler reposted
NVIDIA
@nvidia
Calling all GTC Alumni. #gtc20 online this fall will be the best one yet! Five days. Four continents. Infinite possibilities. Register today: nvda.ws/2ENvzv0
-
Sanja Fidler reposted
Sanyam Bhutani
@bhutanisanyam1
New @ctdsshow Ep:
This is no doubt, one of THE BEST conversations on the show. 🙏
Here's my interview w @dimadamen about:
- 📸 Action & Video Recog
- 🤖 Computer Vision
- 🍛 Epic-Kitchens-100 Dataset
Audio: anchor.fm/chaitimedatascience/episodes/Dima-Damen--Action…
Video: youtu.be/GXqq_hj-UuY -
Sanja Fidler reposted
Alexandre Moufarek
@amoufarek
Procedural models widely used to generate realistic scenes in video games rely on experts to tune a number of parameters. Meta-Sim2 (Devaranjan, Kar, Fidler, 2020) learns to synthesize 3D scenes (structure & parameters) from unlabeled real images.
Paper: arxiv.org/abs/2008.09092 -
Sanja Fidler reposted
Alec Jacobson
@_AlecJacobson
"Nonlinear Color Triads for Approximation, Learning and Direct Manipulation of Color Distributions"
by @arttoolmaker, @amlankar95, @FidlerSanja, @karansher
collaboration with @VectorInst and @nvidia
www.youtube.com/watch?v=GfF4soPVMLw&list=PLIO-p-PoBO8apCe… -
Sanja Fidler reposted
Alec Jacobson
@_AlecJacobson
📢@UofTCompSci presented 8 technical papers at #siggraph2020
UofT authors include
@sellan_s
@HTDerekLi
@arttoolmaker
@rinatgg
@amlankar95
@KazemCH
@MaryamDehnavi
@eitangrinspun
@FidlerSanja
@karansher
David Levin
Talks 🆓publicly🆓 available on youtube
www.youtube.com/playlist?list=PLIO-p-PoBO8apCel7_E3Um87aO… -
Sanja Fidler reposted
Chris Choy, Ph.D.
@realChrisChoy
For those who haven't attended ECCV2020, here's the screenshot of #eccv2020 and ICLR MiniConf.
One of these websites required $1M to implement. -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Check out our #eccv2020 Oral: V2VNet, which leverages vehicle-to-vehicle communication to greatly improve perception and prediction for autonomous driving.
Sessions: 8/26 9am/7pm EST and 8/27 9 pm EST
Paper: arxiv.org/pdf/2008.07519
#selfdrivingcars -
Sanja Fidler reposted
Jonah Philion
@PhilionJonah
"Lift, Splat, Shoot" poster @eccvconf is today! First session begins in 8 hours.
Link: eccv.6connex.eu/event/ECCV2020/
6am-8am / 4pm-6pm San Francisco
8am-10am / 6pm-8pm Chicago
9am-11am / 7pm-9pm New York
2pm-4pm / (Fri) midnight-2am London
9pm-11pm / (Fri) 7am-9am Beijing -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Check this out! How to generate scene-consistent samples in real time for motion forecasting in self-driving. All from raw sensor data while performing perception as well.
simon
@disiok
Check out at #eccv2020 how we generate multiple scene-consistent motion forecasts of complex urban traffic, directly from sensor data!
Paper: arxiv.org/pdf/2007.12036
Live Q&A: Thu 9am & 7pm ET
Joint work with: @sergioksas, @cole_gulino, @katielulula, @lrjconan, @RaquelUrtasun -
Sanja Fidler
@FidlerSanja
Check out our @eccvconf paper on AI-assisted 3D shape creation from images/point clouds. Uses 2D scribbles as input - so no graphics expertise needed to use it. You can try it out on the website!
paper: arxiv.org/abs/2008.10719
web:www.cs.toronto.edu/~shenti11/scribble3d/
@researchuoft @VectorInst
Tianchang Shen
@TianchangS
Excited to present our work “Interactive Annotation of 3D Object Geometry using 2D scribbles” at #eccv2020 tomorrow w/ @JunGao33210520 @amlankar95 @FidlerSanja (time below)!
Given point cloud/imagery, a graphic layman can produce good quality meshes with a few annotations. -
Sanja Fidler
@FidlerSanja
Check out our interactive video segmentation paper at #eccv2020, with a nice performance boost over SOTA.
We'll be doing demos at the poster!
web: cs.toronto.edu/~linghuan/scribblebox/
paper: arxiv.org/abs/2008.09721
Soon in aidemos.cs.toronto.edu/annotation-suite/
@researchuoft @VectorInst
Huan Ling
@HuanLing6
Drop by our @eccvconf paper and see our Demo!!
4287: ScribbleBox: Interactive Annotation Framework for Video Object Segmentation
@FidlerSanja
Session 1: Aug. 27, 2020, 14:00
Session 2: Aug. 28, 2020, 00:00 (UTC+1)
Website: www.cs.toronto.edu/~linghuan/scribblebox/
Arxiv: arxiv.org/abs/2008.09721 -
Sanja Fidler
@FidlerSanja
If you're at @eccvconf, come see Toronto Annotation Suite - our AI-assisted annotation tooling! We've been putting significant effort in building software for dataset labeling, featuring all our AI methods to significantly boost efficiency. @UofTCompSci @researchuoft @VectorInst
Amlan Kar
@amlankar95
We've been putting our AI assisted annotation methods @uoftresearch into a web platform! Come to our demo of the Toronto Annotation Suite at #eccv2020 to try it / get a trial account!
Web: aidemos.cs.toronto.edu/annotation-suite/
When: Tue, Aug 25, 7 PM EDT and Wed, Aug 26, 9 AM EDTThis media couldn't be displayed. See it on the original post.
Sanja Fidler
@FidlerSanja
happening now!
-
Sanja Fidler
@FidlerSanja
From only unlabeled images, MetaSim2 learns to synthesize 3D scenes (structure¶meters) - useful for content creation, gaming, or synthesizing labeled data for computer vision. Check it out at @eccvconf! @amlankar95 @kdbacho @NVIDIAAI @NVIDIAAIDev
web: nv-tlabs.github.io/meta-sim-structure/
Amlan Kar
@amlankar95
Excited to present Meta-Sim2 at #eccv2020 tomorrow! (links below)
We learn to generate synthetic scene structure *unsupervised* from real images by learning to sample rules from a scene grammar
w/ @kdbacho (joint lead (and undergrad!!)), @FidlerSanja at @NVIDIAAI -
Sanja Fidler
@FidlerSanja
Check out our @eccvconf paper Learning Geometric Image Representation with a Deformable Grid - instead of processing images on fixed grids as in traditional pipelines, we learn to deform the grid in end2end manner. Plug&play module, nice accuracy boosts!
arxiv.org/abs/2008.09269
Jun Gao
@JunGao33210520
Drop by our @eccvconf paper!!
2278: Beyond Fixed Grid: Learning Geometric Image Representation with a Deformable Grid @zianwang97 @FidlerSanja
Session 1: Tue 25 Aug 14:00 - 16:00 (now!)
Session 2: Wed 26 Aug 00:00 - 02:00
Website: www.cs.toronto.edu/~jungao/def-grid/
Times UTC+1.
Sanja Fidler
@FidlerSanja
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
Excited to speak on Wednesday at the #selfdrivenwomen panel organized by @Waymo to celebrate #womensequalityday. Join the community of women shaping the future of mobility. This is a field full of opportunities for brilliant people like you!
Waymo
@Waymo
How are you celebrating #womensequalityday on Wednesday, 8/26? Waymo COO @TechTekedra will be joined by @AvaryMichelle, @tilly_chang g & @RaquelUrtasun for our #selfdrivenwomen panel to discuss their careers, the industry, & answer your questions. Sign up: waymoevents.com
-
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
NVIDIA is presenting over 20 papers at #eccv2020. One of them, COCO-FUNIT, is a computer vision model that can translate an image of an animal such as a puppy, and convert it into a snow leopard. See the complete list of papers here: nvda.ws/2Qm89Q8
-
Sanja Fidler reposted
Jan Kautz
@jankautz
Applications are open for our annual NVIDIA Graduate Fellowship Program, seeking doctoral students working in accelerated computing, AI, machine learning, autonomous vehicles, robotics, AI for healthcare, high performance computing and related fields.
blogs.nvidia.com/blog/2020/08/18/gpu-graduate-fellowship/ -
Sanja Fidler
@FidlerSanja
Check out beautiful research led by Masha Shugrina, being presented at SIGGRAPH'20. Come play with #color triads colorsandbox.com! #art @UofTCompSci @researchuoft @VectorInst
Masha Shugrina
@_shumash
Excited to present Color Triads at virtual #siggraph2020 this week: colorsandbox.com/research/triads Joint work with @FidlerSanja, @amlankar95, @karansher at @UofT, @VectorInst.
-
Sanja Fidler
@FidlerSanja
Great work by my student @PhilionJonah on birds-eye-view perception and planning from arbitrary camera rigs for #selfdriving. Code coming soon! @NVIDIAAI @eccvconf
Project page: nv-tlabs.github.io/lift-splat-shoot/
Paper: arxiv.org/abs/2008.05711
Jonah Philion
@PhilionJonah
"Lift, Splat, Shoot: Encoding Images from Arbitrary Camera Rigs by Implicitly Unprojecting to 3D" w/ @FidlerSanja @eccvconf 2020
n images -> bird's-eye-view perception + motion planning
@nvidia @VectorInst @UofTCompSci
nv-tlabs.github.io/lift-splat-shoot/
arxiv.org/abs/2008.05711
(1/6) -
Sanja Fidler reposted
WiCV
@WiCVworkshop
We are pleased to welcome @dimadamen, @FidlerSanja and @zeynepakata as our guest speakers at #wicv. They will be joined by Rahul Sukthankar for a panel discussing diversity in computer vision. Leave your anonymous questions for the panel here: forms.gle/hsxSo5NN3p2S3ujcA
-
Sanja Fidler
@FidlerSanja
Was there any mention of such outrageous publication fee at the time of submission? If I saw it at the time all my submissions would have gone to NeurIPS instead.
Amlan Kar
@amlankar95
This is the last @eccvconf paper I will submit if they continue with Springer. 450 GBP publication fee is more than the price of 20 ICML student registrations.
#eccv2020 #openaccess -
Sanja Fidler
@FidlerSanja
I love the effort of @AI_Scholars_SFU and @ai4allorg for educating young people about AI, focusing on underrepresented groups. I hope to see a lot more diversity in future cohorts of students in college and grad school! AI is cool, come study!
Dr. Angelica Lim @petitegeek.bsky.social
@petitegeek
Thrilled to have Computer Vision expert @FidlerSanja speak to our cohort of @ai4allorg AI scholars this morning. What an inspiration to our 36 young women! Next up is Angel Chang, who will give an intro to NLP tomorrow. Check @AI_Scholars_SFU for the Eventbrite! @CIFAR_News
-
Sanja Fidler reposted
Frank Dellaert
@fdellaert
Please retweet widely, maybe University System of Georgia will respond to media criticism (or is that wishful thinking?)
Mark Riedl
@mark_riedl
Not happy is an understatement www.cnn.com/2020/07/05/us/georgia-tech-reopening-no-masks…
-
Sanja Fidler
@FidlerSanja
My sister writes a beautiful blog about Slovenia’s nature www.exploringslovenia.com. Love these photos. I miss home!! @ExploringSlo
Exploring Slovenia
@ExploringSlo
On the hidden slopes of the Julian Alps lies a remarkable kingdom of the legendary Goldhorn, an Alpine ibex with golden horns that had supposedly chased away everyone intruding upon his territory. Today, the rare Alpine ibex continues the Goldhorn’s famed charisma.
-
Sanja Fidler reposted
Keenan Crane
@keenanisalive
Also, the steering committee and the program chairs for the last 5 years are all men—maybe we should talk about that too?
Alec Jacobson
@_AlecJacobson
All 16 SGP 2020 Graduate School instructors (including me) are men. This extreme lack of gender diversity does not reflect the geometry processing community
Let's talk about diversity at the townhall on Tuesday
meanwhile, let's post great geometry processing papers by women: -
Sanja Fidler
@FidlerSanja
I got tenure at #uoft! Associate professor as of today 😊
-
Sanja Fidler reposted
Sebastian Seung
@SebastianSeung
Starting in July, I’ll be serving as President of @Samsung Research. I look forward to guiding research and development @Samsung, strengthening basic and applied research, and striving for maximum real-world impact.
-
Sanja Fidler reposted
NVIDIA Robotics
@NVIDIARobotics
NVIDIA and Mercedes-Benz AG are taking a giant step forward, delivering vehicle technology unlike anything that's been done before. www.linkedin.com/leap/r?code=dXmeAHS
-
Sanja Fidler reposted
maggie_albrecht
@maggie_albrecht
Meet @FidlerSanja and @jankautz live on Friday June 26th 1pm PDT to learn more about @NVIDIAAI career opportunities.
www.eventbrite.com/e/cvpr-2020-meet-the-managers-live-web… -
Sanja Fidler reposted
Aditya Kusupati
@adityakusupati
After a hectic quarter ending w/ a robotic @icmlconf video recording. Yeah, STR arxiv.org/abs/2002.03231 got in. Excited to explore new things @NVIDIAAI w/ @FidlerSanja & Antonio Torralba.
A great summer in Toronto, Austria & India? - Looking forward to recreating it from Seattle -
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I have uploaded my 9 #cvpr2020 workshop keynotes on my new YouTube channel www.youtube.com/channel/UCLGGmIWc3Ia1jhKBNUsbBjg/. This is 6h of content about all aspects of self-driving: perception, prediction, motion planning, localization, V2V, Q/A, HD maps, etc #selfdrivingcars
-
Sanja Fidler reposted
Jonah Philion
@PhilionJonah
This paper @CVPR shows that neural motion planners induce a salient evaluation metric on perception.
We're releasing code so that other datasets (in addition to nuscenes.org) can easily incorporate our metric if interested.
@NVIDIAAI article: developer.nvidia.com/blog/research-autonomous-vehicle-per…
Sanja Fidler
@FidlerSanja
Excited to release Planning KL - a new task-specific metric for evaluating perception in #selfdriving. @NVIDIAAI We’re collaborating w @Aptiv to include PKL on nuScenes leaderboard. @CVPR Thu 5pm PDT with @PhilionJonah @amlankar95
arxiv.org/abs/2004.08745
nv-tlabs.github.io/detection-relevance/ -
Sanja Fidler
@FidlerSanja
Excited to release Planning KL - a new task-specific metric for evaluating perception in #selfdriving. @NVIDIAAI We’re collaborating w @Aptiv to include PKL on nuScenes leaderboard. @CVPR Thu 5pm PDT with @PhilionJonah @amlankar95
arxiv.org/abs/2004.08745
nv-tlabs.github.io/detection-relevance/ -
Sanja Fidler
@FidlerSanja
Check out our work on auto tuning structured light! @ChenWenzheng @paaarsaaa Kyros Kutulakos #uoft
Parsa Mirdehghan
@paaarsaaa
Excited to share our #cvpr2020 paper, “Auto-Tuning Structured Light by Optical Stochastic Gradient Descent”; joint work with Wenzheng Chen, @FidlerSanja, and Kyros Kutulakos.
Project webpage: dgp.toronto.edu/autotuningsl
Check out our poster sessions: Tue 1-3pm EDT, Wed 1-3am EDT -
Sanja Fidler
@FidlerSanja
cvpr20.com/event/neural-data-server-a-large-scale-search-…
David Acuna
@davidjesusacu
Our work on creating a Machine Learning search engine for Transfer Learning data is an ORAL at #cvpr2020!
Check out the presentation and come to our Q&A session for an Interactive Demo! Tues 4:00 PM - 6:00 PM
@yanxi0830 @FidlerSanja
aidemos.cs.toronto.edu/nds/ -
Sanja Fidler
@FidlerSanja
Join us!
Seung Wook Kim
@seungkim0123
Our work "Learning to Simulate Dynamic Environments with GameGAN" will be presented at CVPR 2020 today!
Check out the project nv-tlabs.github.io/gameGAN/ and come to our Q&A session at 10AM-12PM, 10PM-12AM (PDT) June 16th!
@PhilionJonah @zhouyh7 @FidlerSanja @abtorralba -
Sanja Fidler
@FidlerSanja
Check out our Neural Data Server at CVPR!
David Acuna
@davidjesusacu
Our work on creating a Machine Learning search engine for Transfer Learning data is an ORAL at #cvpr2020!
Check out the presentation and come to our Q&A session for an Interactive Demo! Tues 4:00 PM - 6:00 PM
@yanxi0830 @FidlerSanja
aidemos.cs.toronto.edu/nds/ -
Sanja Fidler reposted
James Tompkin
@jtompkin
AI for Content Creation workshop at #cvpr2020 starts in 15 minutes! visual.cs.brown.edu/aicc2020/ Great list of speakers @phillip_isola @akanazawa @irrfaan @jcjohnss @FidlerSanja @AaronHertzmann ying-cao.com faculty.ucmerced.edu/mhyang/ + audience award for best paper!
-
Sanja Fidler reposted
Vector Institute
@VectorInst
Vector Faculty member @FidlerSanja and co-authors @davidjesusacu & @yanxi0830 will give an oral presentation of her Neural Data Server paper at @CVPR tomorrow.
Read about the paper and more of our #cvpr20 activities: vectorinstitute.ai/2020/06/11/vector-researchers-present-… -
Sanja Fidler reposted
Jan Kautz
@jankautz
And a second tutorial on "Learning Representations via Graph-structured Networks": xiaolonw.github.io/graphnnv2/ (1pm PDT). Join us! @NVIDIAAI (2/2)
-
Sanja Fidler reposted
Jan Kautz
@jankautz
If you are attending #cvpr2020, we are organizing two tutorials today. One on "Novel View Synthesis: From Depth-Based Warping to Multi-Plane Images and Beyond": nvlabs.github.io/nvs-tutorial-cvpr2020/ (9:15am PDT). @NVIDIAAI (1/2)
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
If you are attending #cvpr2020, check out our tutorial on self-driving cars: allaboutselfdriving.com Starting today at 8:45am PT. @CVPR @CVPRConf
-
Sanja Fidler reposted
Raquel Urtasun
@RaquelUrtasun
I finally got a twitter account! @Uber @UofT @VectorInst @UberATG #selfdrivingcars #selfdriving Stay safe!
-
Sanja Fidler reposted
NVIDIA
@nvidia
Trained on 50,000 episodes of PAC-MAN, a new #ai model created by NVIDIA Research called GameGAN re-created the game environment in just days. Celebrate the #pacman40th anniversary and see how GameGAN puts an AI spin on this classic game. nvda.ws/2LPBsZ6 @BandaiNamcoUS
-
Sanja Fidler
@FidlerSanja
Excited to release our new work gameGAN that learns to re-create games by watching screenplay! You'll soon be able to play AI-recreated PacMan on NVIDIA's AI Playground. @seungkim0123 HenryZhou @PhilionJonah @abtorralba
Project page: nv-tlabs.github.io/gameGAN/
NVIDIA AI Developer
@NVIDIAAIDev
Trained on 50,000 episodes of PAC-MAN with @NVIDIA V100 GPUs, this new GAN model developed by NVIDIA Researchers from our Toronto lab, generates a fully functional version of the dot-munching classic without an underlying game engine, just #ai #pacman40th nvda.ws/2LPBsZ6
-
Sanja Fidler reposted
NVIDIA
@nvidia
The #gtc20 Keynote videos are now available. Watch to learn about the latest innovation in #ai, HPC, data science, robotics, graphics and more. nvda.ws/2WtHsfW
-
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
The time is almost here! Tune in this Thursday at 6am pst for @NVIDIA CEO Jensen Huang’s #gtc20 keynote - learn about the latest innovations in #ai, #hpc, robotics, graphics, healthcare, data science, and more. Save the date: nvda.ws/3fFejpJ
-
Sanja Fidler reposted
Bryan Catanzaro
@ctnzr
Jensen has been cooking again! Time to take the world’s largest graphics card out of the oven! Keynote will be live at 6 AM PDT on Thursday, May 14. youtu.be/So7TNRhIYJ8
-
Sanja Fidler
@FidlerSanja
Check out our webinar on Kaolin, 3D Deep Learning Library!
-
Sanja Fidler
@FidlerSanja
Congrats to my amazing PhD student @davidjesusacu for being one of five recipients of the Microsoft Ade Lovelace Fellowship! Very proud of you David! @researchuoft @VectorInst
David Acuna
@davidjesusacu
Extremely proud of being one of the recipients of the 2020 Microsoft Ada Lovelace Fellowship.
-
Sanja Fidler reposted
Amlan Kar
@amlankar95
@PyTorch code showing the generic learning setup and reproducing simple experiments is now available!
Code: github.com/nv-tlabs/meta-sim
Project Page: nv-tlabs.github.io/meta-sim/
Paper: arxiv.org/abs/1904.11621
Sanja Fidler
@FidlerSanja
Excited to release our work on MetaSim which learns to generate synthetic datasets that resemble real datasets in content. @NvidiaAI @amlankar95 @aayush382 @liu_mingyu @davidjesusacu Antonio Torralba etc
paper: arxiv.org/abs/1904.11621
project page: nv-tlabs.github.io/meta-sim/ -
Sanja Fidler reposted
VisualData
@visualdataio
visualdata.io was cited as the major source of computer vision datasets in the new @nvidia paper Neural Data Server authored by @FidlerSanja and her colleagues, where machine learning is used to recommend the best datasets for transfer learning
aidemos.cs.toronto.edu/nds/ -
Sanja Fidler
@FidlerSanja
Current image datasets are getting very large and pre-training on them is time consuming. We are releasing Neural Data Server (NDS), a search engine for transfer learning data! @yanxi0830 @davidjesusacu #uoft
Webservice: aidemos.cs.toronto.edu/nds
Paper: arxiv.org/abs/2001.02799 -
Sanja Fidler
@FidlerSanja
Come and see our Toronto Annotation suite -- annotation powered with A.I. -- demo today at NeurIPS!
Amlan Kar
@amlankar95
We'll be demoing the first version of the Toronto Annotation Suite (aidemos.cs.toronto.edu/annotation-suite/) today at @NeurIPSConf!
Come talk to us at 5:30 PM today at East Exhibition Hall B+C to find out more about cool stuff in AI assisted data labeling and management! -
Sanja Fidler
@FidlerSanja
If you're interested in optical flow, check out Creative Flow+, a large-scale multi-style artistic video dataset (densely labeled with GT)! Awesome job by Masha Shugrina! @arttoolmaker Ziheng Liang @amlankar95 Jiaman Li, Angad Singh, Karan Singh
Webpage: www.cs.toronto.edu/creativeflow/ -
Sanja Fidler reposted
NVIDIA AI
@NVIDIAAI
To help accelerate 3D #deeplearning research, @NVIDIA released Kaolin to provide efficient implementations of differentiable #3d modules. See how it eases users by simply modifying components of the interface supplies.
-
Sanja Fidler
@FidlerSanja
Excited to release a @PyTorch library for 3D deep learning! Check it out, and give us feedback! Great effort by @krrish94 Edward Smith, JF Lafleche @Caenorst Artem Rozantsev, Tommy Xiang, Gav State, @RevLebaredian @NvidiaAI . We plan to extend it with many exciting features!
NVIDIA AI
@NVIDIAAI
Research efforts in #3d computer vision and #ai are on the rise. To accelerate 3D #deeplearning research, NVIDIA releases Kaolin as a PyTorch library. See how researchers use Kaolin to move 3D models into the realm of neural networks. nvda.ws/2OeclzE
Sanja Fidler
@FidlerSanja
Github: github.com/NVIDIAGameWorks/kaolin/
Description: arxiv.org/abs/1911.05063 -
Sanja Fidler reposted
David Acuna
@davidjesusacu
We are presenting Gated-SCNN today at ICCV 19.new SoTA segmentation network that explicitly wires shape into the recognition pipeline. Hall B1, Poster-79. Come and talk to us!!! @yongyuanxi @jampani_varun @FidlerSanja @NvidiaAI
-
Sanja Fidler
@FidlerSanja
Our @NvidiaAI work Meta-Sim on learning to create synthetic datasets presented in today’s morning oral session @ICCV19! Check it out! @amlankar95 @aayush382 @liu_mingyu @davidjesusacu Antonio Torralba @NvidiaAI
Sanja Fidler
@FidlerSanja
Oral session D1
-
Sanja Fidler
@FidlerSanja
Our @NvidiaAI work NTG on city road layout generation will be presented in today’s morning session @ICCV19! Hang Chu, @lidaiqing @davidjesusacu @amlankar95 @liu_mingyu Masha Shugrina,Xinkai Wei,Antonio Torralba
Paper: arxiv.org/abs/1910.02055
Project page: nv-tlabs.github.io/NTG/
Sanja Fidler
@FidlerSanja
Oral session D1
-
Sanja Fidler
@FidlerSanja
Our ICCV work on differentiable bipartite matching (with only a few lines of code!), with application to video segmentation.
Renjie Liao
@lrjconan
Interested in a differentiable bipartite matching algorithm with theoretical convergence guarantee and <50 lines of Pytorch code? Please check our ICCV work:
arxiv.org/abs/1909.12471
Code:
github.com/ZENGXH/DMM_Net
With Xiaohui, Li, Yuwen, @FidlerSanja and Raquel -
Sanja Fidler
@FidlerSanja
Excited to share our @NvidiaAI work on semantic segmentation called GSCNN, which significantly outperforms DeepLabV3+ on Cityscapes benchmark.
@yongyuanxi @davidjesusacu @jampani_varun @NvidiaAI
paper: arxiv.org/abs/1907.05740
project: github.com/nv-tlabs/GSCNN
Code coming soon
Sanja Fidler
@FidlerSanja
to correct link to project page: nv-tlabs.github.io/GSCNN/
:)
Sanja Fidler
@FidlerSanja
Code now released: github.com/nv-tlabs/gscnn
-
Sanja Fidler
@FidlerSanja
We now released code and trained models for our GSCNN semantic segmentation work done at @NvidiaAI. Check it out! @yongyuanxi @davidjesusacu @jampani_varun
Paper (ICCV'19): arxiv.org/abs/1907.05740
Project page: nv-tlabs.github.io/GSCNN/
Code (@PyTorch ): github.com/nv-tlabs/gscnn -
Sanja Fidler reposted
David Acuna
@davidjesusacu
GSCNN accepted at @ICCV19!!!
-
Sanja Fidler reposted
Amlan Kar
@amlankar95
Accepted as an oral at ICCV 2019! :)
Sanja Fidler
@FidlerSanja
Excited to release our work on MetaSim which learns to generate synthetic datasets that resemble real datasets in content. @NvidiaAI @amlankar95 @aayush382 @liu_mingyu @davidjesusacu Antonio Torralba etc
paper: arxiv.org/abs/1904.11621
project page: nv-tlabs.github.io/meta-sim/ -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
NVIDIA breaks 8 #ai training performance records, delivering up to 1.8x faster-time-to-train in new round of #mlperf. Get these #dl software optimizations and models from NGC.
-
Sanja Fidler reposted
maggie_albrecht
@maggie_albrecht
NVIDIA Research Released at CVPR Helps Developers Create Better Visual Datasets - NVIDIA Developer News Center developer.nvidia.com/blog/nvidia-research-released-at-cvp…
-
Sanja Fidler
@FidlerSanja
Releasing VirtualHome, a simulator of complex household activities. VH allows interactions with environment, eg grabbing objects, switching on/off appliances. Marko Boben @xavierpuigf @andrewliao11 @kkhra0 @abtorralba
Page: virtual-home.org
code: github.com/xavierpuigf/virtualhome
Sanja Fidler
@FidlerSanja
Poster and demo of VirtualHome now at CVPR: Poster Session 2-1P, poster 206, 10am-12.45pm. Marko Boben @xavierpuigf @andrewliao11 @kkhra0 @abtorralba #cvpr19
-
Sanja Fidler
@FidlerSanja
We are demoing our interactive annotation methods at posters at CVPR today. Come check it out! @HuanLing6 @JunGao33210520 @amlankar95 Zian Wang @davidjesusacu #cvpr2019
Curve-GCN: session 2-1P, poster 71, 10am-12.45pm
Deep Levelset: session 2-2P, poster 77, 3.20pm-6pm
Sanja Fidler
@FidlerSanja
Curve-GCN: arxiv.org/abs/1903.06874
Deep Levelset: openaccess.thecvf.com/content_CVPR_2019/papers/Wang_Objec… -
Sanja Fidler
@FidlerSanja
NVIDIA Research will present 20 papers at #cvpr2019, including 11 orals. Full list is here: nvda.ws/31xiZab. Check it out!
-
Sanja Fidler
@FidlerSanja
Releasing STEAL, a new semantic boundary detector that significantly outperforms past work. Use STEAL to refine segmentation datasets, and train better segmentation models! @davidjesusacu @amlankar95 @NvidiaAI
paper:arxiv.org/abs/1904.07934
code(@PyTorch):github.com/nv-tlabs/STEAL
Sanja Fidler
@FidlerSanja
Check out David Acuna @davidjesusacu give a talk at CVPR next week: 1330 – 1520, Oral session 3.2C
-
Post not available here, view on X
Sanja Fidler
@FidlerSanja
Check out David Acuna @davidjesusacu give a talk at CVPR next week: Thu, 1330-1520, Oral session 3.2C
-
Sanja Fidler reposted
Guodong Zhang
@Guodzh
Excited to share a new paper with Chaoqi, @RogerGrosse and @FidlerSanja: EigenDamage: Structured Pruning in the Kronecker-Factored Eigenbasis, appearing at @icmlconf.
Paper: arxiv.org/abs/1905.05934
Code: github.com/alecwangcq/EigenDamage-Pytorch -
Sanja Fidler
@FidlerSanja
Excited to release our work on MetaSim which learns to generate synthetic datasets that resemble real datasets in content. @NvidiaAI @amlankar95 @aayush382 @liu_mingyu @davidjesusacu Antonio Torralba etc
paper: arxiv.org/abs/1904.11621
project page: nv-tlabs.github.io/meta-sim/
Sanja Fidler
@FidlerSanja
-
Sanja Fidler
@FidlerSanja
We released our new interactive annotation approach, which outperforms Polygon-RNN++ and is 10x faster. Great work by @HuanLing6, @JunGao33210520, @amlankar95, @ChenWenzheng!
Paper: arxiv.org/pdf/1903.06874.pdf
Video: www.youtube.com/watch?v=ycD2BtO-QzU
Code (@PyTorch): github.com/fidler-lab/curve-gcn -
Sanja Fidler
@FidlerSanja
NVIDIA CEO Jensen Huang delivered a sweeping opening keynote to kick off #gtc19, describing the company’s progress accelerating the sprawling datacenters that power the world’s most dynamic industries. Missed it? Check out the recap here: nvda.ws/2OccuU1
-
Sanja Fidler reposted
NVIDIA Robotics
@NVIDIARobotics
We just launched the 3rd AI City Challenge and looking for teams to participate. Learn more about how you can push the boundaries for AI-based traffic optimization. nvda.ws/2C1S2zx
This media couldn't be displayed. See it on the original post. -
Sanja Fidler
@FidlerSanja
Thanks to CIFAR for the Canadian AI Chair award!
-
Sanja Fidler reposted
NVIDIA
@nvidia
NVIDIA has set six records in AI, demonstrating the unmatched performance and versatility of NVIDIA Tensor Core GPUs and driving innovation for the evolution of AI. nvda.ws/2LbWM9Y #mlperf
-
Sanja Fidler
@FidlerSanja
Our latest work on vision and language:
Lifelong Learning for Image Captioning by Asking Natural Language Questions, arxiv.org/abs/1812.00235
by Kevin Shen & @amlankar95
Visual Reasoning by Progressive Module Networks, arxiv.org/abs/1806.02453
by @seungkim0123 & Makarand Tapaswi -
Sanja Fidler reposted
NVIDIA AI Developer
@NVIDIAAIDev
Meet with NVIDIA's #ai and #deeplearning experts at #neurips2018. Stop by booth 209 to check out office hours Monday - Thursday. nvda.ws/2Bi4ETS
-
Sanja Fidler reposted
Clement Farabet
@clmt
Pretty cool to see NVIDIA up this list. Def an amazing place to work and innovate. www.forbes.com/sites/justcapital/2018/08/31/just-capitals…
-
Sanja Fidler
@FidlerSanja
We just released code and labeling demo for our interactive annotation tool PolygonRnn++: github.com/fidler-lab/polyrnn-pp-pytorch. Great work and effort by @davidjesusacu , Huan Ling, and @amlankar95 !









































































































































