Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Meet LA-Pose. Our latest model taking Wayve another step towards generalization at scale. LA-Pose employs large-scale self-supervised learning, building strong motion representations for 3D perception from 10.2 million unlabeled driving video snippets, unlike today's strongest approaches that often depend on expensive, carefully curated 3D supervision. With only a lightweight...

wayve_ai's profile picture

Wayve

16,785 subscribers

36,869 Aufrufe • vor 3 Monaten •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Meta presents Sapiens Foundation for Human Vision Models discuss: We present Sapiens, a family of models for four fundamental human-centric vision tasks - 2D pose estimation, body-part segmentation, depth estimation, and surface normal prediction. Our models natively support 1K high-resolution inference and are extremely easy to adapt for individual tasks by simply fine-tuning models pretrained on over 300 million in-the-wild human images. We observe that, given the same computational budget, self-supervised pretraining on a curated dataset of human images significantly boosts the performance for a diverse set of human-centric tasks. The resulting models exhibit remarkable generalization to in-the-wild data, even when labeled data is scarce or entirely synthetic. Our simple model design also brings scalability - model performance across tasks improves as we scale the number of parameters from 0.3 to 2 billion. Sapiens consistently surpasses existing baselines across various human-centric benchmarks. We achieve significant improvements over the prior state-of-the-art on Humans-5K (pose) by 7.6 mAP, Humans-2K (part-seg) by 17.1 mIoU, Hi4D (depth) by 22.4% relative RMSE, and THuman2 (normal) by 53.5% relative angular error.
2:00

Meta presents Sapiens Foundation for Human Vision Models discuss: We present Sapiens, a family of models for four fundamental human-centric vision tasks - 2D pose estimation, body-part segmentation, depth estimation, and surface normal prediction. Our models natively support 1K high-resolution inference and are extremely easy to adapt for individual tasks by simply fine-tuning models pretrained on over 300 million in-the-wild human images. We observe that, given the same computational budget, self-supervised pretraining on a curated dataset of human images significantly boosts the performance for a diverse set of human-centric tasks. The resulting models exhibit remarkable generalization to in-the-wild data, even when labeled data is scarce or entirely synthetic. Our simple model design also brings scalability - model performance across tasks improves as we scale the number of parameters from 0.3 to 2 billion. Sapiens consistently surpasses existing baselines across various human-centric benchmarks. We achieve significant improvements over the prior state-of-the-art on Humans-5K (pose) by 7.6 mAP, Humans-2K (part-seg) by 17.1 mIoU, Hi4D (depth) by 22.4% relative RMSE, and THuman2 (normal) by 53.5% relative angular error.

AK

151,614 Aufrufe • vor 2 Jahren

Nvidia announces GAvatar: Animatable 3D Gaussian Avatars with Implicit Mesh Learning paper page: Gaussian splatting has emerged as a powerful 3D representation that harnesses the advantages of both explicit (mesh) and implicit (NeRF) 3D representations. In this paper, we seek to leverage Gaussian splatting to generate realistic animatable avatars from textual descriptions, addressing the limitations (e.g., flexibility and efficiency) imposed by mesh or NeRF-based representations. However, a naive application of Gaussian splatting cannot generate high-quality animatable avatars and suffers from learning instability; it also cannot capture fine avatar geometries and often leads to degenerate body parts. To tackle these problems, we first propose a primitive-based 3D Gaussian representation where Gaussians are defined inside pose-driven primitives to facilitate animation. Second, to stabilize and amortize the learning of millions of Gaussians, we propose to use neural implicit fields to predict the Gaussian attributes (e.g., colors). Finally, to capture fine avatar geometries and extract detailed meshes, we propose a novel SDF-based implicit mesh learning approach for 3D Gaussians that regularizes the underlying geometries and extracts highly detailed textured meshes. Our proposed method, GAvatar, enables the large-scale generation of diverse animatable avatars using only text prompts. GAvatar significantly surpasses existing methods in terms of both appearance and geometry quality, and achieves extremely fast rendering (100 fps) at 1K resolution.
5:36

Nvidia announces GAvatar: Animatable 3D Gaussian Avatars with Implicit Mesh Learning paper page: Gaussian splatting has emerged as a powerful 3D representation that harnesses the advantages of both explicit (mesh) and implicit (NeRF) 3D representations. In this paper, we seek to leverage Gaussian splatting to generate realistic animatable avatars from textual descriptions, addressing the limitations (e.g., flexibility and efficiency) imposed by mesh or NeRF-based representations. However, a naive application of Gaussian splatting cannot generate high-quality animatable avatars and suffers from learning instability; it also cannot capture fine avatar geometries and often leads to degenerate body parts. To tackle these problems, we first propose a primitive-based 3D Gaussian representation where Gaussians are defined inside pose-driven primitives to facilitate animation. Second, to stabilize and amortize the learning of millions of Gaussians, we propose to use neural implicit fields to predict the Gaussian attributes (e.g., colors). Finally, to capture fine avatar geometries and extract detailed meshes, we propose a novel SDF-based implicit mesh learning approach for 3D Gaussians that regularizes the underlying geometries and extracts highly detailed textured meshes. Our proposed method, GAvatar, enables the large-scale generation of diverse animatable avatars using only text prompts. GAvatar significantly surpasses existing methods in terms of both appearance and geometry quality, and achieves extremely fast rendering (100 fps) at 1K resolution.

AK

140,992 Aufrufe • vor 2 Jahren

🚨 Paper Alert 🚨 ➡️Paper Title: Articulate3D: Zero-Shot Text-Driven 3D Object Posing 🌟Few pointers from the paper 🎯Authors of this paper proposed a training-free method, “Articulate3D”, to pose a 3D asset through language control. 🎯Despite advances in vision and language models, this task remains surprisingly challenging. 🎯To achieve this goal, they decomposed the problem into two steps. 🎯They modified a powerful image-generator to create target images conditioned on the input image and a text instruction. 🎯They then align the mesh to the target images through a multi-view pose optimisation step. 🎯 In detail, they introduced a self-attention rewiring mechanism (RSActrl) that decouples the source structure from pose within an image generative model, allowing it to maintain a consistent structure across varying poses. 🎯They observed that differentiable rendering is an unreliable signal for articulation optimisation; instead, they used keypoints to establish correspondences between input and target images. 🎯The effectiveness of Articulate3D is demonstrated across a diverse range of 3D objects and free-form text prompts, successfully manipulating poses while maintaining the original identity of the mesh. 🎯Quantitative evaluations and a comparative user study, in which their method was preferred over 85% of the time, confirm its superiority over existing approaches. 🏢Organization: University of Oxford , Google DeepMind 🧙Paper Authors: Oishi Deb, Anjun Hu, Ashkan Khakzar, Philip Torr, Christian Rupprecht 📝 Read the Full Paper here: 🗂️ Project Page: 🎥 Be sure to watch the attached Demo Video - Sound on 🔊🔊 Find this Valuable 💎 ? ♻️QT and teach your network something new Follow me 👣, naveen manwani , for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements.
0:46

🚨 Paper Alert 🚨 ➡️Paper Title: Articulate3D: Zero-Shot Text-Driven 3D Object Posing 🌟Few pointers from the paper 🎯Authors of this paper proposed a training-free method, “Articulate3D”, to pose a 3D asset through language control. 🎯Despite advances in vision and language models, this task remains surprisingly challenging. 🎯To achieve this goal, they decomposed the problem into two steps. 🎯They modified a powerful image-generator to create target images conditioned on the input image and a text instruction. 🎯They then align the mesh to the target images through a multi-view pose optimisation step. 🎯 In detail, they introduced a self-attention rewiring mechanism (RSActrl) that decouples the source structure from pose within an image generative model, allowing it to maintain a consistent structure across varying poses. 🎯They observed that differentiable rendering is an unreliable signal for articulation optimisation; instead, they used keypoints to establish correspondences between input and target images. 🎯The effectiveness of Articulate3D is demonstrated across a diverse range of 3D objects and free-form text prompts, successfully manipulating poses while maintaining the original identity of the mesh. 🎯Quantitative evaluations and a comparative user study, in which their method was preferred over 85% of the time, confirm its superiority over existing approaches. 🏢Organization: University of Oxford , Google DeepMind 🧙Paper Authors: Oishi Deb, Anjun Hu, Ashkan Khakzar, Philip Torr, Christian Rupprecht 📝 Read the Full Paper here: 🗂️ Project Page: 🎥 Be sure to watch the attached Demo Video - Sound on 🔊🔊 Find this Valuable 💎 ? ♻️QT and teach your network something new Follow me 👣, naveen manwani , for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements.

naveen manwani

14,334 Aufrufe • vor 1 Jahr

Love seeing Silico (Goodfire ) used to probe our EchoJEPA's representations! this is exactly the kind of interpretability work that's been missing for JEPA-style models. One thing that makes EchoJEPA particularly interesting to interpret: unlike MAE-based approaches, it never reconstructs pixels. The model learns entirely in latent space through masked prediction, so you can't just look at decoder outputs to understand what it captured. Attribution onto a temporally aligned 3D mesh is a much more honest probe of what the representations actually encode. What we found in building EchoJEPA: training on 18M echo videos across 300K patients, the model learns to disentangle cardiac anatomy from ultrasound noise (speckle, reverberation artifacts) almost entirely through self-supervision. With 1% labeled data it already outperforms supervised baselines trained on 100%. The latent space is doing real anatomical work, but until you can visualize it like this, "real anatomical work" is mostly a claim. Paper + code: |
0:51

Love seeing Silico (Goodfire ) used to probe our EchoJEPA's representations! this is exactly the kind of interpretability work that's been missing for JEPA-style models. One thing that makes EchoJEPA particularly interesting to interpret: unlike MAE-based approaches, it never reconstructs pixels. The model learns entirely in latent space through masked prediction, so you can't just look at decoder outputs to understand what it captured. Attribution onto a temporally aligned 3D mesh is a much more honest probe of what the representations actually encode. What we found in building EchoJEPA: training on 18M echo videos across 300K patients, the model learns to disentangle cardiac anatomy from ultrasound noise (speckle, reverberation artifacts) almost entirely through self-supervision. With 1% labeled data it already outperforms supervised baselines trained on 100%. The latent space is doing real anatomical work, but until you can visualize it like this, "real anatomical work" is mostly a claim. Paper + code: |

Bo Wang

29,611 Aufrufe • vor 3 Monaten

A new 30-minute presentation from Ashok Elluswamy, Tesla’s VP of AI, has been released, where he talks about FSD, AI and the team’s latest progress. Highlight from the presentation: • Tesla's vehicle fleet can provide 500 years of driving data every single day. Curse of Dimensionality: • 8 cameras at high frame rate = billions of tokens per 30 seconds of driving context. • Tesla must compress and extract the right correlations between sensory input and control actions. Data Advantage: • Tesla has access to a “Niagara Falls of data” — hundreds of years’ worth of collective fleet driving. • Uses smart data triggers to capture rare corner cases (e.g., complex intersections, unpredictable behavior). Quality and Efficiency: • Extracts only the essential data needed to train models efficiently. Debugging and Interpretability: • Even though the system is end-to-end, Tesla can still prompt the model to output interpretable data: 3D occupancy, road boundaries, objects, signs, traffic lights, etc. • Natural language querying: ask the model why it made a certain decision. • These auxiliary predictions don’t drive the car but help engineers debug and ensure safety. Tesla’s Advanced Gaussian Splatting (3D Scene Modeling): • Tesla developed a custom, ultra-fast Gaussian splatting system to reconstruct 3D scenes from limited camera views. • Produces crisp, accurate 3D renderings even from few camera angles — far better than standard NeRF/splatting approaches. • Enables rapid visual debugging of the driving environment in 3D. Evaluation & World Models: • Evaluation is the hardest challenge: models may perform well offline but fail in real-world conditions. • Tesla builds balanced, diverse evaluation datasets focusing on edge cases — not just easy highway driving. Introduced a learned world simulator (neural network-generated video engine): • Can simulate 8 Tesla camera feeds simultaneously — fully synthetic. • Used for testing, training, and reinforcement learning. • Allows adversarial event injection (e.g., adding a pedestrian or vehicle cutting in). • Enables replaying past failures to verify new model improvements. • Can run in near real-time, letting testers “drive” inside a simulated world. What’s Next: • Scale robotaxi service globally. • Unlock full autonomy across the entire Tesla fleet. • Cybercab: next-gen 2-seat vehicle designed specifically for robotaxi use, targeting lowest transportation cost (cheaper than public transit). • Same neural networks will power Optimus humanoid robot. • The same video generation system is now being applied to Optimus. • The system can simulate and plan movement for robots, adapting easily to new forms. via the International Conference on Computer Vision (ICCV). Full presentation:
0:29

A new 30-minute presentation from Ashok Elluswamy, Tesla’s VP of AI, has been released, where he talks about FSD, AI and the team’s latest progress. Highlight from the presentation: • Tesla's vehicle fleet can provide 500 years of driving data every single day. Curse of Dimensionality: • 8 cameras at high frame rate = billions of tokens per 30 seconds of driving context. • Tesla must compress and extract the right correlations between sensory input and control actions. Data Advantage: • Tesla has access to a “Niagara Falls of data” — hundreds of years’ worth of collective fleet driving. • Uses smart data triggers to capture rare corner cases (e.g., complex intersections, unpredictable behavior). Quality and Efficiency: • Extracts only the essential data needed to train models efficiently. Debugging and Interpretability: • Even though the system is end-to-end, Tesla can still prompt the model to output interpretable data: 3D occupancy, road boundaries, objects, signs, traffic lights, etc. • Natural language querying: ask the model why it made a certain decision. • These auxiliary predictions don’t drive the car but help engineers debug and ensure safety. Tesla’s Advanced Gaussian Splatting (3D Scene Modeling): • Tesla developed a custom, ultra-fast Gaussian splatting system to reconstruct 3D scenes from limited camera views. • Produces crisp, accurate 3D renderings even from few camera angles — far better than standard NeRF/splatting approaches. • Enables rapid visual debugging of the driving environment in 3D. Evaluation & World Models: • Evaluation is the hardest challenge: models may perform well offline but fail in real-world conditions. • Tesla builds balanced, diverse evaluation datasets focusing on edge cases — not just easy highway driving. Introduced a learned world simulator (neural network-generated video engine): • Can simulate 8 Tesla camera feeds simultaneously — fully synthetic. • Used for testing, training, and reinforcement learning. • Allows adversarial event injection (e.g., adding a pedestrian or vehicle cutting in). • Enables replaying past failures to verify new model improvements. • Can run in near real-time, letting testers “drive” inside a simulated world. What’s Next: • Scale robotaxi service globally. • Unlock full autonomy across the entire Tesla fleet. • Cybercab: next-gen 2-seat vehicle designed specifically for robotaxi use, targeting lowest transportation cost (cheaper than public transit). • Same neural networks will power Optimus humanoid robot. • The same video generation system is now being applied to Optimus. • The system can simulate and plan movement for robots, adapting easily to new forms. via the International Conference on Computer Vision (ICCV). Full presentation:

Sawyer Merritt

1,287,671 Aufrufe • vor 10 Monaten

I was really impressed by the UMI gripper (Cheng Chi et al.), but a key limitation is that **force-related data wasn’t captured**: humans feel haptic feedback through the mechanical springs, but the robot couldn’t leverage that info, limiting the data’s value for fine-grained manipulation tasks. Led by my amazing students Yolanda Zhu and Binghao Huang, we designed a **portable visuo-tactile gripper** by integrating our dense, flexible tactile arrays with the UMI gripper to enable large-scale in-the-wild data collection. 🔗 We demonstrate **cross-modal representation learning** and **downstream policy learning** on tasks requiring in-hand state estimation (e.g., test tube reorientation) and fine-grained force sensing (e.g., pipette fluid transfer). Key takeaways: - Our flexible tactile arrays store the rich haptic information humans perceive as dense tactile signals. - Portability and robustness are key for in-the-wild data collection; our portable gripper is compact, lightweight, and durable. - Touch provides precise, robust measurements of in-hand object pose, invariant to lighting and viewpoint. - Cross-modal pretraining on large-scale in-the-wild data significantly improves policy robustness and sample efficiency (as shown many times before — and verified again here!). Also check out our previous investigations of dense, flexible tactile grids for understanding human-robot-environment interactions: - Dense tactile glove (Nature ’19): - 3D-ViTac (CoRL ’24):
5:35

I was really impressed by the UMI gripper (Cheng Chi et al.), but a key limitation is that **force-related data wasn’t captured**: humans feel haptic feedback through the mechanical springs, but the robot couldn’t leverage that info, limiting the data’s value for fine-grained manipulation tasks. Led by my amazing students Yolanda Zhu and Binghao Huang, we designed a **portable visuo-tactile gripper** by integrating our dense, flexible tactile arrays with the UMI gripper to enable large-scale in-the-wild data collection. 🔗 We demonstrate **cross-modal representation learning** and **downstream policy learning** on tasks requiring in-hand state estimation (e.g., test tube reorientation) and fine-grained force sensing (e.g., pipette fluid transfer). Key takeaways: - Our flexible tactile arrays store the rich haptic information humans perceive as dense tactile signals. - Portability and robustness are key for in-the-wild data collection; our portable gripper is compact, lightweight, and durable. - Touch provides precise, robust measurements of in-hand object pose, invariant to lighting and viewpoint. - Cross-modal pretraining on large-scale in-the-wild data significantly improves policy robustness and sample efficiency (as shown many times before — and verified again here!). Also check out our previous investigations of dense, flexible tactile grids for understanding human-robot-environment interactions: - Dense tactile glove (Nature ’19): - 3D-ViTac (CoRL ’24):

Yunzhu Li

13,188 Aufrufe • vor 1 Jahr

Physics-based Motion Retargeting from Sparse Inputs paper page: Avatars are important to create interactive and immersive experiences in virtual worlds. One challenge in animating these characters to mimic a user's motion is that commercial AR/VR products consist only of a headset and controllers, providing very limited sensor data of the user's pose. Another challenge is that an avatar might have a different skeleton structure than a human and the mapping between them is unclear. In this work we address both of these challenges. We introduce a method to retarget motions in real-time from sparse human sensor data to characters of various morphologies. Our method uses reinforcement learning to train a policy to control characters in a physics simulator. We only require human motion capture data for training, without relying on artist-generated animations for each avatar. This allows us to use large motion capture datasets to train general policies that can track unseen users from real and sparse data in real-time. We demonstrate the feasibility of our approach on three characters with different skeleton structure: a dinosaur, a mouse-like creature and a human. We show that the avatar poses often match the user surprisingly well, despite having no sensor information of the lower body available. We discuss and ablate the important components in our framework, specifically the kinematic retargeting step, the imitation, contact and action reward as well as our asymmetric actor-critic observations. We further explore the robustness of our method in a variety of settings including unbalancing, dancing and sports motions.
7:12

Physics-based Motion Retargeting from Sparse Inputs paper page: Avatars are important to create interactive and immersive experiences in virtual worlds. One challenge in animating these characters to mimic a user's motion is that commercial AR/VR products consist only of a headset and controllers, providing very limited sensor data of the user's pose. Another challenge is that an avatar might have a different skeleton structure than a human and the mapping between them is unclear. In this work we address both of these challenges. We introduce a method to retarget motions in real-time from sparse human sensor data to characters of various morphologies. Our method uses reinforcement learning to train a policy to control characters in a physics simulator. We only require human motion capture data for training, without relying on artist-generated animations for each avatar. This allows us to use large motion capture datasets to train general policies that can track unseen users from real and sparse data in real-time. We demonstrate the feasibility of our approach on three characters with different skeleton structure: a dinosaur, a mouse-like creature and a human. We show that the avatar poses often match the user surprisingly well, despite having no sensor information of the lower body available. We discuss and ablate the important components in our framework, specifically the kinematic retargeting step, the imitation, contact and action reward as well as our asymmetric actor-critic observations. We further explore the robustness of our method in a variety of settings including unbalancing, dancing and sports motions.

AK

106,527 Aufrufe • vor 3 Jahren

Today we're announcing #GAIA1: a 9B parameter world model, trained on 4,700 hours of driving data, able to simulate complex and diverse driving scenes from video, text and action inputs. This model is 480x larger than the preview we shared earlier this year and the results are incredible. These videos are entirely synthetically generated by Wayve's generative AI, GAIA-1. But there is more here than just generating videos, GAIA is an entire world model. A world model allows us to simulate the future, conditioned on video, text and action inputs, which can be leveraged for making informed decisions when driving. Why is this game-changing for autonomous driving? 1. Safety. One limitation with AI systems like today's Large Language Models is that they are autoregressive, next-word prediction algorithms, but aren't necessarily aware of the implications of their decisions. A world model allows us to give our AI the capability to be aware of its decisions, by simulating the future, which is important for self-driving safety. 2. Synthetic training data. I believe synthetic training data is the future for AI, because it is safer, cheaper, and infinitely scalable. GAIA-1 unlocks unprecedented realism and diversity of synthetic data for self-driving. 3. Long-tail robustness. One of the biggest challenges for self-driving is long-tail robustness: dealing with the enormous magnitude of edge cases we see on the road. An advantage of generative AI is its incredible ability to recombine experiences in new ways. This is exciting for self-driving as it means we can learn from two edge case scenarios, and combine them to become a corner case. For example, we can experience driving in fog, and experience of jay-walking pedestrians, and GAIA can learn from these experiences to understand how to generate a fog+jay walking scenario. Check out many more videos in our blog or further technical details in our paper: Or come chat with our team who are at the International Conference on Computer Vision (#ICCV2023) this week in Paris in Booth 32 Jamie Shotton
0:41

Today we're announcing #GAIA1: a 9B parameter world model, trained on 4,700 hours of driving data, able to simulate complex and diverse driving scenes from video, text and action inputs. This model is 480x larger than the preview we shared earlier this year and the results are incredible. These videos are entirely synthetically generated by Wayve's generative AI, GAIA-1. But there is more here than just generating videos, GAIA is an entire world model. A world model allows us to simulate the future, conditioned on video, text and action inputs, which can be leveraged for making informed decisions when driving. Why is this game-changing for autonomous driving? 1. Safety. One limitation with AI systems like today's Large Language Models is that they are autoregressive, next-word prediction algorithms, but aren't necessarily aware of the implications of their decisions. A world model allows us to give our AI the capability to be aware of its decisions, by simulating the future, which is important for self-driving safety. 2. Synthetic training data. I believe synthetic training data is the future for AI, because it is safer, cheaper, and infinitely scalable. GAIA-1 unlocks unprecedented realism and diversity of synthetic data for self-driving. 3. Long-tail robustness. One of the biggest challenges for self-driving is long-tail robustness: dealing with the enormous magnitude of edge cases we see on the road. An advantage of generative AI is its incredible ability to recombine experiences in new ways. This is exciting for self-driving as it means we can learn from two edge case scenarios, and combine them to become a corner case. For example, we can experience driving in fog, and experience of jay-walking pedestrians, and GAIA can learn from these experiences to understand how to generate a fog+jay walking scenario. Check out many more videos in our blog or further technical details in our paper: Or come chat with our team who are at the International Conference on Computer Vision (#ICCV2023) this week in Paris in Booth 32 Jamie Shotton

Alex Kendall

631,869 Aufrufe • vor 2 Jahren

If you want a vision encoder for dexterous manipulation, what should be the most important part to model? 🤔 Current standard models like CLIP, SigLIP, and DINOv2 have an incredible grasp of semantics and spatial details. But they lack the action-centric structure needed for downstream visuomotor control. But collecting annotated robotic trajectories at scale is SUPER expensive and largely unrealistic. So, how do we bridge this gap? We introduce CAIP (Contrastive Action-Image Pre-training) ⬇️ 🔸 Action-centric upstream: we align visual observations with action chunks through a contrastive objective. 🔸 Human video as a proxy: we represent 3D human hand poses analogously to robotic end-effector actions, tapping into a massive source of human demonstrations. 🔸 Massive scale: pre-trained on over 32,000 hours of manipulation video, driving both sample efficiency and robust generalization. 🔸 Hardware proven: achieves a 76% average success rate on a real-world Dexmate Vega bimanual Dexmate manipulator with dual 22-DoF Sharpa Wave hands Sharpa . 🔸 State-of-the-art: significantly outperforms strong baselines like DINOv2, SigLIP, MVP, and Qwen3.5 ViT across complex tasks, even under unexpected lighting changes and visual distractors. 🌐 Project: 📝 Blog: 📄 Paper: 💻 Model:
1:30

If you want a vision encoder for dexterous manipulation, what should be the most important part to model? 🤔 Current standard models like CLIP, SigLIP, and DINOv2 have an incredible grasp of semantics and spatial details. But they lack the action-centric structure needed for downstream visuomotor control. But collecting annotated robotic trajectories at scale is SUPER expensive and largely unrealistic. So, how do we bridge this gap? We introduce CAIP (Contrastive Action-Image Pre-training) ⬇️ 🔸 Action-centric upstream: we align visual observations with action chunks through a contrastive objective. 🔸 Human video as a proxy: we represent 3D human hand poses analogously to robotic end-effector actions, tapping into a massive source of human demonstrations. 🔸 Massive scale: pre-trained on over 32,000 hours of manipulation video, driving both sample efficiency and robust generalization. 🔸 Hardware proven: achieves a 76% average success rate on a real-world Dexmate Vega bimanual Dexmate manipulator with dual 22-DoF Sharpa Wave hands Sharpa . 🔸 State-of-the-art: significantly outperforms strong baselines like DINOv2, SigLIP, MVP, and Qwen3.5 ViT across complex tasks, even under unexpected lighting changes and visual distractors. 🌐 Project: 📝 Blog: 📄 Paper: 💻 Model:

Roei Herzig

23,239 Aufrufe • vor 2 Monaten

We are excited to share our work “Event-Aided Sharp Radiance Field Reconstruction for Fast-Flying Drones” published in IEEE Transactions on Robotics IEEE Transactions on Robotics (T-RO), which tackles sharp radiance field reconstruction under agile drone motion, where RGB frames are heavily motion-blurred and pose priors become unreliable! 4 years in the making! Code & dataset released! PDF: Code & Dataset: Full Narrated Video: High-speed flight is essential for time- and battery-constrained missions (e.g., inspection, exploration, search & rescue). However, fast motion corrupts visual data with severe motion blur and introduces drift/noise in visual-inertial odometry, making NeRF-based 3D reconstruction particularly brittle. We propose a unified framework that leverages asynchronous #EventCamera streams together with motion-blurred frames to reconstruct high-fidelity radiance fields from agile drone flights. Our key idea is to embed event-image fusion directly into radiance field optimization while jointly refining a shared, continuous-time camera trajectory initialized from event-based VIO. This enables us to recover sharp radiance fields and accurate trajectories without ground-truth supervision during training. We validate our method on synthetic data and on real sequences captured by a drone flying up to 2 m/s. Despite severe blur and noisy pose priors, our method preserves fine scene details and achieves a performance gain of over 50% on real-world data compared to state-of-the-art methods. Kudos to Rong Zou and Marco Cannici! Marco Cannici Reference: Rong Zou*, Marco Cannici*, Davide Scaramuzza Event-Aided Sharp Radiance Field Reconstruction for Fast-Flying Drones IEEE Transactions on Robotics (T-RO), 2026 NCCR Robotics European Research Council (ERC) AUTOASSESS UZH IfI University of Zurich UZH Science Prophesee SynSense UZH Space Hub
1:02

We are excited to share our work “Event-Aided Sharp Radiance Field Reconstruction for Fast-Flying Drones” published in IEEE Transactions on Robotics IEEE Transactions on Robotics (T-RO), which tackles sharp radiance field reconstruction under agile drone motion, where RGB frames are heavily motion-blurred and pose priors become unreliable! 4 years in the making! Code & dataset released! PDF: Code & Dataset: Full Narrated Video: High-speed flight is essential for time- and battery-constrained missions (e.g., inspection, exploration, search & rescue). However, fast motion corrupts visual data with severe motion blur and introduces drift/noise in visual-inertial odometry, making NeRF-based 3D reconstruction particularly brittle. We propose a unified framework that leverages asynchronous #EventCamera streams together with motion-blurred frames to reconstruct high-fidelity radiance fields from agile drone flights. Our key idea is to embed event-image fusion directly into radiance field optimization while jointly refining a shared, continuous-time camera trajectory initialized from event-based VIO. This enables us to recover sharp radiance fields and accurate trajectories without ground-truth supervision during training. We validate our method on synthetic data and on real sequences captured by a drone flying up to 2 m/s. Despite severe blur and noisy pose priors, our method preserves fine scene details and achieves a performance gain of over 50% on real-world data compared to state-of-the-art methods. Kudos to Rong Zou and Marco Cannici! Marco Cannici Reference: Rong Zou*, Marco Cannici*, Davide Scaramuzza Event-Aided Sharp Radiance Field Reconstruction for Fast-Flying Drones IEEE Transactions on Robotics (T-RO), 2026 NCCR Robotics European Research Council (ERC) AUTOASSESS UZH IfI University of Zurich UZH Science Prophesee SynSense UZH Space Hub

Davide Scaramuzza

12,028 Aufrufe • vor 5 Monaten

🚨 SIGGRAPH Asia 2025 Paper Alert 🚨 ➡️Paper Title: WorldExplorer: Towards Generating Fully Navigable 3D Scenes 🌟Few pointers from the paper 🎯Generating 3D worlds from text is a highly anticipated goal in computer vision. Existing works are limited by the degree of exploration they allow inside of a scene, i.e., produce stretched-out and noisy artifacts when moving beyond central or panoramic perspectives. 🎯 To this end, authors of this paper proposed “WorldExplorer”, a novel method based on autoregressive video trajectory generation, which builds fully navigable 3D scenes with consistent visual quality across a wide range of viewpoints. 🎯They initialize their scenes by creating multi-view consistent images corresponding to a 360 degree panorama. 🎯Then, they expanded it by leveraging video diffusion models in an iterative scene generation pipeline. 🎯Concretely, they generated multiple videos along short, pre-defined trajectories, that explore the scene in depth, including motion around objects. 🎯Their novel scene memory conditions each video on the most relevant prior views, while a collision-detection mechanism prevents degenerate results, like moving into objects. 🎯Finally,they fuse all generated views into a unified 3D representation via 3D Gaussian Splatting optimization. 🎯Compared to prior approaches, WorldExplorer produces high-quality scenes that remain stable under large camera motion, enabling for the first time realistic and unrestricted exploration. 🎯They believe this marks a significant step toward generating immersive and truly explorable virtual 3D environments. 🏢Organization: TU München 🧙Paper Authors: Manuel-Andreas Schneider, Lukas Höllein , Matthias Niessner 📝 Read the Full Paper here: 🗂️ Project Page: 🧑‍💻 Code: 🎥 Be sure to watch the attached Technical Summary Video - Sound on 🔊🔊 Find this Valuable 💎 ? ♻️QT and teach your network something new Follow me 👣, naveen manwani , for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. #SIGGRAPHAsia2025
3:36

🚨 SIGGRAPH Asia 2025 Paper Alert 🚨 ➡️Paper Title: WorldExplorer: Towards Generating Fully Navigable 3D Scenes 🌟Few pointers from the paper 🎯Generating 3D worlds from text is a highly anticipated goal in computer vision. Existing works are limited by the degree of exploration they allow inside of a scene, i.e., produce stretched-out and noisy artifacts when moving beyond central or panoramic perspectives. 🎯 To this end, authors of this paper proposed “WorldExplorer”, a novel method based on autoregressive video trajectory generation, which builds fully navigable 3D scenes with consistent visual quality across a wide range of viewpoints. 🎯They initialize their scenes by creating multi-view consistent images corresponding to a 360 degree panorama. 🎯Then, they expanded it by leveraging video diffusion models in an iterative scene generation pipeline. 🎯Concretely, they generated multiple videos along short, pre-defined trajectories, that explore the scene in depth, including motion around objects. 🎯Their novel scene memory conditions each video on the most relevant prior views, while a collision-detection mechanism prevents degenerate results, like moving into objects. 🎯Finally,they fuse all generated views into a unified 3D representation via 3D Gaussian Splatting optimization. 🎯Compared to prior approaches, WorldExplorer produces high-quality scenes that remain stable under large camera motion, enabling for the first time realistic and unrestricted exploration. 🎯They believe this marks a significant step toward generating immersive and truly explorable virtual 3D environments. 🏢Organization: TU München 🧙Paper Authors: Manuel-Andreas Schneider, Lukas Höllein , Matthias Niessner 📝 Read the Full Paper here: 🗂️ Project Page: 🧑‍💻 Code: 🎥 Be sure to watch the attached Technical Summary Video - Sound on 🔊🔊 Find this Valuable 💎 ? ♻️QT and teach your network something new Follow me 👣, naveen manwani , for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. #SIGGRAPHAsia2025

naveen manwani

10,578 Aufrufe • vor 11 Monaten

I am happy to be finally able to post what I was able to build over the last few weeks. A full real-time high-frequency state estimation and mapping algorithm completely written line by line from scratch in Rust, which can be used by robots to navigate and reason within the 3D world also in complicated scenarios. TBH this took me longer than expected (which was still super fast :D) but you need to get a lot right: From the sensors over the drivers to their respective estimation pipeline and then fusing everything together - a covariance nightmare - and something that can be refined over years to come (currently using Fisher Information from the real measurements). What you see here is not the output of some structure from motion or Gaussian splatting, these are the points of a tight mesh (high res for the video) that a robot can use in real time to plan a path using any open-source planner. The flight you experience through the world is the actual state estimate of the scanner which is published at IMU rate. Yes, currently we have some artefacts of filtered-out humans (GDPR compliant of course :) ) and moving cars and there is still some calibration that could be improved. Offline refinement with SFM and Gaussian splats is possible as well but currently not on the road map. What is on the road map is an exciting step of now being able to collect data from customers at construction sites and in warehouses (currently handheld in the near future with a robot). This data can then be used by our physical agents to reason within this world and automate any customer’s task related to 3D data. If you have anyone who wastes time manually looking 👀 through 3D data, or cannot collect enough 3D data and interpret: Tell me how to reach them!
1:15

I am happy to be finally able to post what I was able to build over the last few weeks. A full real-time high-frequency state estimation and mapping algorithm completely written line by line from scratch in Rust, which can be used by robots to navigate and reason within the 3D world also in complicated scenarios. TBH this took me longer than expected (which was still super fast :D) but you need to get a lot right: From the sensors over the drivers to their respective estimation pipeline and then fusing everything together - a covariance nightmare - and something that can be refined over years to come (currently using Fisher Information from the real measurements). What you see here is not the output of some structure from motion or Gaussian splatting, these are the points of a tight mesh (high res for the video) that a robot can use in real time to plan a path using any open-source planner. The flight you experience through the world is the actual state estimate of the scanner which is published at IMU rate. Yes, currently we have some artefacts of filtered-out humans (GDPR compliant of course :) ) and moving cars and there is still some calibration that could be improved. Offline refinement with SFM and Gaussian splats is possible as well but currently not on the road map. What is on the road map is an exciting step of now being able to collect data from customers at construction sites and in warehouses (currently handheld in the near future with a robot). This data can then be used by our physical agents to reason within this world and automate any customer’s task related to 3D data. If you have anyone who wastes time manually looking 👀 through 3D data, or cannot collect enough 3D data and interpret: Tell me how to reach them!

Benedikt Seidel

16,656 Aufrufe • vor 3 Monaten

NVIDIA just unleashed SANA-WM and it’s an absolute MONSTER for the future of open source AI! A blazing-fast 2.6B-parameter open-source world model that doesn’t just generate video… it creates controllable, physics-rich, high-fidelity worlds on demand. Why this is insanely powerful: • One image + text prompt + 6-DoF camera trajectory → generates 720p videos up to 60 seconds long with buttery-smooth, precisely controlled camera movement. You’re not just watching, you’re piloting the simulation. • Runs locally on a single consumer GPU (RTX 5090 level) thanks to heavy distillation + NVFP4 quantization. Full 60-second clip denoised in ~34 seconds. No massive clusters required. • 36× higher throughput than previous open models while rivaling (or beating) closed industrial giants in visual quality and consistency. • Trained lightning-fast: ~213K public videos in just 15 days on 64 H100s. • Built with next-level tech: Hybrid Linear Attention, dual-branch camera control, two-stage pipeline, and rock-solid metric-scale pose understanding. This is a true open world model, the foundation for embodied AI, robotics, autonomous systems, and hyper-realistic simulations that can run anywhere. Project: At our Zero-Human Company, we’re already running SANA-WM live in our core pipelines. It’s supercharging autonomous agent training, generating unlimited synthetic training data, and powering full end-to-end simulation loops, zero humans in the loop. The speed and control let us test thousands of edge-case scenarios overnight, iterate at lightspeed, and push our fully autonomous operations further than ever before. This is the kind of breakthrough that turns science fiction into daily reality. World models just leveled up — hard. The age of personal, local, controllable universes is here.
1:13

NVIDIA just unleashed SANA-WM and it’s an absolute MONSTER for the future of open source AI! A blazing-fast 2.6B-parameter open-source world model that doesn’t just generate video… it creates controllable, physics-rich, high-fidelity worlds on demand. Why this is insanely powerful: • One image + text prompt + 6-DoF camera trajectory → generates 720p videos up to 60 seconds long with buttery-smooth, precisely controlled camera movement. You’re not just watching, you’re piloting the simulation. • Runs locally on a single consumer GPU (RTX 5090 level) thanks to heavy distillation + NVFP4 quantization. Full 60-second clip denoised in ~34 seconds. No massive clusters required. • 36× higher throughput than previous open models while rivaling (or beating) closed industrial giants in visual quality and consistency. • Trained lightning-fast: ~213K public videos in just 15 days on 64 H100s. • Built with next-level tech: Hybrid Linear Attention, dual-branch camera control, two-stage pipeline, and rock-solid metric-scale pose understanding. This is a true open world model, the foundation for embodied AI, robotics, autonomous systems, and hyper-realistic simulations that can run anywhere. Project: At our Zero-Human Company, we’re already running SANA-WM live in our core pipelines. It’s supercharging autonomous agent training, generating unlimited synthetic training data, and powering full end-to-end simulation loops, zero humans in the loop. The speed and control let us test thousands of edge-case scenarios overnight, iterate at lightspeed, and push our fully autonomous operations further than ever before. This is the kind of breakthrough that turns science fiction into daily reality. World models just leveled up — hard. The age of personal, local, controllable universes is here.

Brian Roemmele

619,016 Aufrufe • vor 3 Monaten

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data. 2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the *motion* of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro -> RoboCasa produces N (varying visuals) -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are building tools to enable everyone in the ecosystem to scale up with us. Links in thread:
1:21

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data. 2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the *motion* of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro -> RoboCasa produces N (varying visuals) -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are building tools to enable everyone in the ecosystem to scale up with us. Links in thread:

Jim Fan

364,565 Aufrufe • vor 2 Jahren

Introducing LifeGPT, showing that LLMs can simulate complex, Turing-complete systems like Conway's Game of Life with near-perfect accuracy—no prior topology needed.🌐This unlocks new potential for AI in modeling self-organizing systems in biology, materials science, & beyond.🔬🤖 #AI #LifeGPT. Cellular Automata (CA), like Conway's Game of Life ("Life"), are computationally irreducible, meaning their evolution is difficult to predict without an a-priori understanding of the rules of the game, including the topology on which it is played. LifeGPT is a topology-agnostic generative model that learns the rules of Life without prior knowledge of its grid structure or boundary conditions, from only a tiny number of game states. The success in simulating Life suggests promising avenues for scientific discovery, particularly in bridging the gap between AI, artificial life, and real-world biological systems, for both forward and inverse problems. The potential for universal computation within generative AI, including LLMs, through approaches like LifeGPT, represents an exciting area for future research, especially when combined with reinforcement learning. Model Convergence: LifeGPT exhibits rapid convergence during training, achieving high accuracy in predicting next-game-states. We attribute the non-zero cross-entropy loss to the lack of causal relationships within randomly generated ICs. Accuracy & Temperature: LifeGPT achieves near-perfect accuracy, particularly at lower sampling temperatures, but can be continually tuned towards higher creativity to discover patterns that the original ruleset would not be able to produce. This finding highlights the trade-off between model creativity (higher temperature) and accuracy in deterministic predictions, with high relevance to model real-world dynamical systems for which no closed-form rulesets exist. Zero/Few-Shot Learning: Trained on a small fraction of possible initial conditions, LifeGPT demonstrates strong zero/few-shot learning, accurately simulating Life for unseen initial conditions. However, rare prediction errors highlight that LifeGPT approximates rather than perfectly replicates the Life algorithm. Autoregressive Autoregressor: A recursive implementation of LifeGPT demonstrates the model's ability to simulate Life over multiple timesteps. LifeGPT is topology-agnostic with respect to its training data and our results show that a GPT model is capable of capturing the deterministic rules of a Turing-complete system with near-perfect accuracy, given sufficiently diverse training data. The work showcases the possibility for future models to synthesize stochastic generative capabilities with deterministic computational capabilities. Link to code, paper, etc. below. Podcast generated using #NotebookLM. LAMM@MIT DMSE at MIT
9:59

Introducing LifeGPT, showing that LLMs can simulate complex, Turing-complete systems like Conway's Game of Life with near-perfect accuracy—no prior topology needed.🌐This unlocks new potential for AI in modeling self-organizing systems in biology, materials science, & beyond.🔬🤖 #AI #LifeGPT. Cellular Automata (CA), like Conway's Game of Life ("Life"), are computationally irreducible, meaning their evolution is difficult to predict without an a-priori understanding of the rules of the game, including the topology on which it is played. LifeGPT is a topology-agnostic generative model that learns the rules of Life without prior knowledge of its grid structure or boundary conditions, from only a tiny number of game states. The success in simulating Life suggests promising avenues for scientific discovery, particularly in bridging the gap between AI, artificial life, and real-world biological systems, for both forward and inverse problems. The potential for universal computation within generative AI, including LLMs, through approaches like LifeGPT, represents an exciting area for future research, especially when combined with reinforcement learning. Model Convergence: LifeGPT exhibits rapid convergence during training, achieving high accuracy in predicting next-game-states. We attribute the non-zero cross-entropy loss to the lack of causal relationships within randomly generated ICs. Accuracy & Temperature: LifeGPT achieves near-perfect accuracy, particularly at lower sampling temperatures, but can be continually tuned towards higher creativity to discover patterns that the original ruleset would not be able to produce. This finding highlights the trade-off between model creativity (higher temperature) and accuracy in deterministic predictions, with high relevance to model real-world dynamical systems for which no closed-form rulesets exist. Zero/Few-Shot Learning: Trained on a small fraction of possible initial conditions, LifeGPT demonstrates strong zero/few-shot learning, accurately simulating Life for unseen initial conditions. However, rare prediction errors highlight that LifeGPT approximates rather than perfectly replicates the Life algorithm. Autoregressive Autoregressor: A recursive implementation of LifeGPT demonstrates the model's ability to simulate Life over multiple timesteps. LifeGPT is topology-agnostic with respect to its training data and our results show that a GPT model is capable of capturing the deterministic rules of a Turing-complete system with near-perfect accuracy, given sufficiently diverse training data. The work showcases the possibility for future models to synthesize stochastic generative capabilities with deterministic computational capabilities. Link to code, paper, etc. below. Podcast generated using #NotebookLM. LAMM@MIT DMSE at MIT

Markus J. Buehler

114,237 Aufrufe • vor 1 Jahr

New model: your robot can now pack your suitcase 🧳 Xiaomi has released a new robot foundation model. Called Xiaomi-Robotics-1, it is designed to have a robot pick things up and move them around. But first, DEFINITIONS: - Mixture-of-Transformers (MoT): An architecture where separate transformer "experts" (e.g., one for vision-language, one for actions) share a single attention stream, so each modality gets specialized parameters without losing joint reasoning. - Vision-language model (VLM): A model that jointly understands images and text. - Diffusion transformer: A transformer trained to turn noise into structured outputs by iterative denoising, here generating robot actions rather than images. - Action chunks: Short sequences of future actions (e.g., the next ~50 motor commands) predicted in one shot instead of one step at a time. - Flow matching: A faster version of diffusion. The model learns a straight-line velocity field from noise to the target action, so it needs only a few integration steps instead of many denoising ones. Its peculiarity comes from its two stage training: 1. 100,000 hours of video shot through a UMI rig: a handheld 3D-printed gripper with a camera, worn by humans doing ordinary tasks in homes, shops, factories and offices. 2. Adapt to actual robot bodies with ~10,000 hours of real-robot data. It replaces the standard approach of teleoperating a real robot for every hour of training data. Its architecture is a Mixture-of-Transformers pairing a pre-trained Qwen3-VL vision-language model with a diffusion transformer that emits action chunks via flow matching, released in 2.6B, 5.1B and 10.5B parameter variants. However, if you read the entire paper ("Scaling VLA Models with over 100K Hours"), you realize that all of the scaling experiments on 20k hours. Therefore the headline "out-of-the-box success climbing 26% → 75% as pre-training data grows" tops out at 100% of 20k hours! What the full corpus does to that curve is never shown -> and this where things would become interesting! Xiaomi's own conclusion is that model size has stopped mattering and data is the binding constraint. The performance gap among different model sizes are less pronounced than those observed across different data scales. This result suggests that model capacity at the billions-parameter scale may already be sufficient to capture the current dataset's distribution. Which further asks the same question: why not use the 100k video hours? Anyway, I would definitely love to have a couple robots at home that can cooperate to pack my suitcase with items relevant to my next destination:
0:58

New model: your robot can now pack your suitcase 🧳 Xiaomi has released a new robot foundation model. Called Xiaomi-Robotics-1, it is designed to have a robot pick things up and move them around. But first, DEFINITIONS: - Mixture-of-Transformers (MoT): An architecture where separate transformer "experts" (e.g., one for vision-language, one for actions) share a single attention stream, so each modality gets specialized parameters without losing joint reasoning. - Vision-language model (VLM): A model that jointly understands images and text. - Diffusion transformer: A transformer trained to turn noise into structured outputs by iterative denoising, here generating robot actions rather than images. - Action chunks: Short sequences of future actions (e.g., the next ~50 motor commands) predicted in one shot instead of one step at a time. - Flow matching: A faster version of diffusion. The model learns a straight-line velocity field from noise to the target action, so it needs only a few integration steps instead of many denoising ones. Its peculiarity comes from its two stage training: 1. 100,000 hours of video shot through a UMI rig: a handheld 3D-printed gripper with a camera, worn by humans doing ordinary tasks in homes, shops, factories and offices. 2. Adapt to actual robot bodies with ~10,000 hours of real-robot data. It replaces the standard approach of teleoperating a real robot for every hour of training data. Its architecture is a Mixture-of-Transformers pairing a pre-trained Qwen3-VL vision-language model with a diffusion transformer that emits action chunks via flow matching, released in 2.6B, 5.1B and 10.5B parameter variants. However, if you read the entire paper ("Scaling VLA Models with over 100K Hours"), you realize that all of the scaling experiments on 20k hours. Therefore the headline "out-of-the-box success climbing 26% → 75% as pre-training data grows" tops out at 100% of 20k hours! What the full corpus does to that curve is never shown -> and this where things would become interesting! Xiaomi's own conclusion is that model size has stopped mattering and data is the binding constraint. The performance gap among different model sizes are less pronounced than those observed across different data scales. This result suggests that model capacity at the billions-parameter scale may already be sufficient to capture the current dataset's distribution. Which further asks the same question: why not use the 100k video hours? Anyway, I would definitely love to have a couple robots at home that can cooperate to pack my suitcase with items relevant to my next destination:

Léo

15,049 Aufrufe • vor 18 Tagen

Everyone in Embodied AI is talking about Vision-Language-Action (VLA) models. Almost no one is talking about the physical nightmare of collecting the data to train them. You can't scrape a kitchen table or a warehouse shelf from a web browser. To get to millions of hours of diverse, real-world manipulation data, you need hundreds of rigs. But you can't buy them. If you build them out of off-the-shelf developer kits, they weigh 15 pounds, overheat in an hour, require bulky cabling, and break the first time an operator wears them on a job site. At the Instawork Robotics Lab (IRL), we had to build our own. Meet the Instacore: a rugged, 4lb wearable egocentric data-capture engine designed specifically to survive a full shift on a standard power pack. We didn't build a flashy humanoid. We did hard, blue-collar systems engineering: 💾 THE COMPUTE — A custom MediaTek Genio carrier board that runs completely fanless, routing 5 camera streams directly to on-board storage. ⚡ THE I/O PIPELINE — We ditched USB for industrial GMSL. Thin, ultra-flexible coaxial lines route high-speed data down to the backpack and Power-over-Coax (PoC) back up, completely eliminating batteries on the wrist. ⏱️ UNIFIED SENSOR CLOCK — The MediaTek Genio SoC drives a shared master clock straight to the ISPs driving our 5 global shutter sensors, stamping metadata at the microsecond of capture to ensure zero-drift temporal alignment. 👁️ OPTIMIZED OPTICS — 95 DFOV lenses on flexible PCB ribbon modules keep the wrist cams flat to prevent snags. A 145-degree chest camera captures the macro workspace, while a 50mm baseline rectilinear stereo head pair preserves close-up 3D mapping. To build hundreds of these, we took over a warehouse in Mountain View in April, called it the Instalab, and brought in talented Pros with assembly backgrounds from Tesla, Apple, and Meta. To test the systems, we had our Pros wear active rigs while assembling more rigs. The video below shows the raw, time-synchronized Foxglove playback of that exact loop. Now, we’re shipping these units globally to scale data collection for real Pros on the job.
1:19

Everyone in Embodied AI is talking about Vision-Language-Action (VLA) models. Almost no one is talking about the physical nightmare of collecting the data to train them. You can't scrape a kitchen table or a warehouse shelf from a web browser. To get to millions of hours of diverse, real-world manipulation data, you need hundreds of rigs. But you can't buy them. If you build them out of off-the-shelf developer kits, they weigh 15 pounds, overheat in an hour, require bulky cabling, and break the first time an operator wears them on a job site. At the Instawork Robotics Lab (IRL), we had to build our own. Meet the Instacore: a rugged, 4lb wearable egocentric data-capture engine designed specifically to survive a full shift on a standard power pack. We didn't build a flashy humanoid. We did hard, blue-collar systems engineering: 💾 THE COMPUTE — A custom MediaTek Genio carrier board that runs completely fanless, routing 5 camera streams directly to on-board storage. ⚡ THE I/O PIPELINE — We ditched USB for industrial GMSL. Thin, ultra-flexible coaxial lines route high-speed data down to the backpack and Power-over-Coax (PoC) back up, completely eliminating batteries on the wrist. ⏱️ UNIFIED SENSOR CLOCK — The MediaTek Genio SoC drives a shared master clock straight to the ISPs driving our 5 global shutter sensors, stamping metadata at the microsecond of capture to ensure zero-drift temporal alignment. 👁️ OPTIMIZED OPTICS — 95 DFOV lenses on flexible PCB ribbon modules keep the wrist cams flat to prevent snags. A 145-degree chest camera captures the macro workspace, while a 50mm baseline rectilinear stereo head pair preserves close-up 3D mapping. To build hundreds of these, we took over a warehouse in Mountain View in April, called it the Instalab, and brought in talented Pros with assembly backgrounds from Tesla, Apple, and Meta. To test the systems, we had our Pros wear active rigs while assembling more rigs. The video below shows the raw, time-synchronized Foxglove playback of that exact loop. Now, we’re shipping these units globally to scale data collection for real Pros on the job.

Ryan Hickman

26,385 Aufrufe • vor 2 Monaten

GRAPHENE GOES INDUSTRIAL: WHY HYDROGRAPH'S PROCESS IS REDEFINING WHAT GRAPHENE CAN BE Most people see graphene as a promising lab curiosity with limited commercial use. Kjirstin Breure, CEO of HydroGraph, saw it as the foundation for upgrading materials across every major industry. She focused on solving the hardest problems in production and proof. The results have created one of the fastest value creations in advanced materials. THE GRAPHENE PRODUCTION EDGE ➡️ Graphene is a single atomic layer of carbon atoms. It stands as the strongest and most conductive material ever discovered. ➡️ Traditional approaches start with graphite and struggle to reach true single-layer quality at scale. HydroGraph uses a detonation synthesis process that converts hydrocarbon gases directly into graphene in one efficient step. ➡️ This method requires no reliance on Chinese graphite supplies and offers theoretically unlimited production capacity with lower capital intensity. THE DATA THAT SEPARATES WINNERS FROM HYPE ➡️ Breure knew data would be everything. HydroGraph has built extensive application development results across concrete, carbon fiber, plastics, and other materials. ➡️ The company supplied large volumes of graphene early to the Graphene Engineering Innovation Centre in Manchester. Independent testing confirmed superior performance that competitors could not match. ➡️ She has been clear about the outcome. "We have proven that we have the highest purity, the highest consistency, and very likely the only genuine graphene at least at industrial scale in the industry." THE INFLECTION POINTS AHEAD ➡️ Since taking the CEO role in March 2024, Breure executed a full business model overhaul. Securing pipeline access for acetylene gas removed the biggest production constraint. ➡️ Company valuation rose from roughly $20 million to about $1 billion in roughly 18 months of disciplined execution. ➡️ The next phase is accelerating fast. A new Austin headquarters, Texas production facility, US redomiciling, NASDAQ listing, and growing defense sector partnerships are all in motion. THE BOTTOM LINE Kjirstin Breure has proven that graphene's prospects are not about hype but about who can deliver real purity, consistency, and scale. HydroGraph is now positioned to lead that charge into commercial reality. The strongest material on earth is finally getting the production platform it deserves. #Graphene #HydroGraph #KjirstinBreure #AdvancedMaterials #MaterialsInnovation #DefenseTech #IndustrialScale
1:17

GRAPHENE GOES INDUSTRIAL: WHY HYDROGRAPH'S PROCESS IS REDEFINING WHAT GRAPHENE CAN BE Most people see graphene as a promising lab curiosity with limited commercial use. Kjirstin Breure, CEO of HydroGraph, saw it as the foundation for upgrading materials across every major industry. She focused on solving the hardest problems in production and proof. The results have created one of the fastest value creations in advanced materials. THE GRAPHENE PRODUCTION EDGE ➡️ Graphene is a single atomic layer of carbon atoms. It stands as the strongest and most conductive material ever discovered. ➡️ Traditional approaches start with graphite and struggle to reach true single-layer quality at scale. HydroGraph uses a detonation synthesis process that converts hydrocarbon gases directly into graphene in one efficient step. ➡️ This method requires no reliance on Chinese graphite supplies and offers theoretically unlimited production capacity with lower capital intensity. THE DATA THAT SEPARATES WINNERS FROM HYPE ➡️ Breure knew data would be everything. HydroGraph has built extensive application development results across concrete, carbon fiber, plastics, and other materials. ➡️ The company supplied large volumes of graphene early to the Graphene Engineering Innovation Centre in Manchester. Independent testing confirmed superior performance that competitors could not match. ➡️ She has been clear about the outcome. "We have proven that we have the highest purity, the highest consistency, and very likely the only genuine graphene at least at industrial scale in the industry." THE INFLECTION POINTS AHEAD ➡️ Since taking the CEO role in March 2024, Breure executed a full business model overhaul. Securing pipeline access for acetylene gas removed the biggest production constraint. ➡️ Company valuation rose from roughly $20 million to about $1 billion in roughly 18 months of disciplined execution. ➡️ The next phase is accelerating fast. A new Austin headquarters, Texas production facility, US redomiciling, NASDAQ listing, and growing defense sector partnerships are all in motion. THE BOTTOM LINE Kjirstin Breure has proven that graphene's prospects are not about hype but about who can deliver real purity, consistency, and scale. HydroGraph is now positioned to lead that charge into commercial reality. The strongest material on earth is finally getting the production platform it deserves. #Graphene #HydroGraph #KjirstinBreure #AdvancedMaterials #MaterialsInnovation #DefenseTech #IndustrialScale

Mark

28,793 Aufrufe • vor 2 Monaten

Open science is how we continue to push technology forward and today at Meta FAIR we’re sharing eight new AI research artifacts including new models, datasets and code to inspire innovation in the community. More in the video from Joelle Pineau. This work is another important step towards our goal of achieving Advanced Machine Intelligence (AMI). What we’re releasing: • Meta Spirit LM: An open source language model for seamless speech and text integration. • Meta Segment Anything Model 2.1: An updated checkpoint with improved results on visually similar objects, small objects and occlusion handling. Plus a new developer suite to make it easier for developers to build with SAM 2. • Layer Skip: Inference code and fine-tuned checkpoints demonstrating a new method for enhancing LLM performance. • SALSA: New code to enable researchers to benchmark AI-based attacks in support of validating security for post-quantum cryptography. • Meta Lingua: A lightweight and self-contained codebase designed to train language models at scale. • Meta Open Materials: New open source models and the largest dataset of its kind to accelerate AI-driven discovery of new inorganic materials. • MEXMA: A new research paper and code for our novel pre-trained cross-lingual sentence encoder with coverage across 80 languages. • Self-Taught Evaluator: a new method for generating synthetic preference data to train reward models without relying on human annotations. Access to state-of-the-art AI creates opportunities for everyone. We’re excited to share this work and look forward to seeing the community innovation that results from it. Details and access to everything released by FAIR today ➡️
1:36

Open science is how we continue to push technology forward and today at Meta FAIR we’re sharing eight new AI research artifacts including new models, datasets and code to inspire innovation in the community. More in the video from Joelle Pineau. This work is another important step towards our goal of achieving Advanced Machine Intelligence (AMI). What we’re releasing: • Meta Spirit LM: An open source language model for seamless speech and text integration. • Meta Segment Anything Model 2.1: An updated checkpoint with improved results on visually similar objects, small objects and occlusion handling. Plus a new developer suite to make it easier for developers to build with SAM 2. • Layer Skip: Inference code and fine-tuned checkpoints demonstrating a new method for enhancing LLM performance. • SALSA: New code to enable researchers to benchmark AI-based attacks in support of validating security for post-quantum cryptography. • Meta Lingua: A lightweight and self-contained codebase designed to train language models at scale. • Meta Open Materials: New open source models and the largest dataset of its kind to accelerate AI-driven discovery of new inorganic materials. • MEXMA: A new research paper and code for our novel pre-trained cross-lingual sentence encoder with coverage across 80 languages. • Self-Taught Evaluator: a new method for generating synthetic preference data to train reward models without relying on human annotations. Access to state-of-the-art AI creates opportunities for everyone. We’re excited to share this work and look forward to seeing the community innovation that results from it. Details and access to everything released by FAIR today ➡️

AI at Meta

150,406 Aufrufe • vor 1 Jahr

We're excited to launch 🚀Airtable AI Assistant 🚀 today, along with AI document analysis and AI web research capabilities! Airtable was founded 12 years ago with the mission of democratizing software creation. Our pioneering innovation was to distill app-building concepts (data, logic, interface) into intuitive visual components, like a no-code lego kit for app building. At the time, we speculated that someday, maybe AI would get good enough to enable conversational app building–talking to an expert AI app builder–and be a huge unlock, making app building even more accessible. We’re now at that point. While surprisingly impressive text generation and manipulation by LLMs was the breakthrough of the 2022 ChatGPT moment, the emergence of surprisingly impressive reasoning capability from LLMs is the breakthrough of 2025. This is unlocking more autonomous agentic experiences, and generating apps and code is the first killer use case (Cursor, @windsurf, Devin, v0, bolt.new, Replit ⠕ Agent to name a few). But for the large class of non-technical builders, a different approach is needed. When AI generates apps with code, rather than no-code building blocks, it requires a developer to fully understand how they work – and to verify them for hidden mistakes that would be tricky/impossible to debug by interface inspection alone (it may look right, but what is the data model business logic is flawed in non-obvious ways?). Airtable Assistant is an agent that can build and modify Airtable apps through conversation, changing schemas, adding automations, and designing interfaces. You can ask it to do things like: –“Research every conference attendee in this base” to have the Assistant immediately spin up an army of researchers that pull in background information for your attendees –“Analyze each contract to identify key risks they pose to my business” to have the Assistant add an AI field that runs an analysis at scale for each contract you’ve signed. Airtable Assistant can also answer questions about the data in your apps, like prompts as advanced as: –“I'm about to meet Jane Smith at Zelos, read all of their recent sales call transcripts and tell me how far along they are in their implementation and if they’re dealing with any issues” –“What are the most common risk factors in our contracts? Are there any changes to our default posture we might consider?” Credit to Mike Krieger for introducing us to the concept of low floor and high ceiling in HCI many years ago, which has become part of our internal lexicon for thinking about product improvements. Assistant dramatically lowers the floor to building apps, including more sophisticated ones, by helping human builders translate their business requirements into the schema design, logic, and interfaces required to deliver on the use case. In addition to launching Airtable Assistant today, we’re also releasing the capability to deploy thousands of AI web researchers, and AI document analysts, to continuously work on the data in Airtable apps. You can do things like: –Pull strategy and value stories from every product requirement doc to draft launch and release messaging –Monitor all brand mentions across digital channels to measure campaign impact –Create an automatically updating competitive intelligence dossier with the latest news and messaging from every competitor in your industry Check it out 👇
1:43

We're excited to launch 🚀Airtable AI Assistant 🚀 today, along with AI document analysis and AI web research capabilities! Airtable was founded 12 years ago with the mission of democratizing software creation. Our pioneering innovation was to distill app-building concepts (data, logic, interface) into intuitive visual components, like a no-code lego kit for app building. At the time, we speculated that someday, maybe AI would get good enough to enable conversational app building–talking to an expert AI app builder–and be a huge unlock, making app building even more accessible. We’re now at that point. While surprisingly impressive text generation and manipulation by LLMs was the breakthrough of the 2022 ChatGPT moment, the emergence of surprisingly impressive reasoning capability from LLMs is the breakthrough of 2025. This is unlocking more autonomous agentic experiences, and generating apps and code is the first killer use case (Cursor, @windsurf, Devin, v0, bolt.new, Replit ⠕ Agent to name a few). But for the large class of non-technical builders, a different approach is needed. When AI generates apps with code, rather than no-code building blocks, it requires a developer to fully understand how they work – and to verify them for hidden mistakes that would be tricky/impossible to debug by interface inspection alone (it may look right, but what is the data model business logic is flawed in non-obvious ways?). Airtable Assistant is an agent that can build and modify Airtable apps through conversation, changing schemas, adding automations, and designing interfaces. You can ask it to do things like: –“Research every conference attendee in this base” to have the Assistant immediately spin up an army of researchers that pull in background information for your attendees –“Analyze each contract to identify key risks they pose to my business” to have the Assistant add an AI field that runs an analysis at scale for each contract you’ve signed. Airtable Assistant can also answer questions about the data in your apps, like prompts as advanced as: –“I'm about to meet Jane Smith at Zelos, read all of their recent sales call transcripts and tell me how far along they are in their implementation and if they’re dealing with any issues” –“What are the most common risk factors in our contracts? Are there any changes to our default posture we might consider?” Credit to Mike Krieger for introducing us to the concept of low floor and high ceiling in HCI many years ago, which has become part of our internal lexicon for thinking about product improvements. Assistant dramatically lowers the floor to building apps, including more sophisticated ones, by helping human builders translate their business requirements into the schema design, logic, and interfaces required to deliver on the use case. In addition to launching Airtable Assistant today, we’re also releasing the capability to deploy thousands of AI web researchers, and AI document analysts, to continuously work on the data in Airtable apps. You can do things like: –Pull strategy and value stories from every product requirement doc to draft launch and release messaging –Monitor all brand mentions across digital channels to measure campaign impact –Create an automatically updating competitive intelligence dossier with the latest news and messaging from every competitor in your industry Check it out 👇

Howie Liu

2,313,248 Aufrufe • vor 1 Jahr