🤯🤯🤯

Introducing Diffusion Forcing, which unifies next-token prediction (eg LLMs) and full-seq. diffusion (eg SORA)! It offers improved performance & new sampling strategies in vision and robotics, such as stable, infinite video generation, better diffusion planning, and more! (1/8)

Academia is catching up fast on humanoid research. To put it in perspective, this is: A team of 5 building with off-the-shelf hardware likely <200k in funding a few months of work. What's even better is all results are open-sourced. The capital efficiency is mind-blowing.

Check out this work from Mohit Shridhar !

I’ve been training dogs since middle school. It’s about time I train robot dogs too 😛 Introducing, UMI on Legs, an approach for scaling manipulation skills on robot dogs🐶It can toss, push heavy weights, and make your ~existing~ visuo-motor policies mobile!

Incredible progress with imitation learning; the learning from demos efficiency is pretty wild

End-to-end imitation for surgical robots! Careful action space design continue to play a key role in getting real-world systems to work

Yesterday, I recorded 100 trajectories to teach my robot arm to grasp LEGO block Overnight, I trained a neural network on my MacBookPro. Today, it can grasp pretty well. No need for an expensive setup to get started with robotics. You just need github.com/huggingface/le…

When I started my first project on in-hand manipulation, I thought it would be super cool but also quite challenging to make my robot hands spin pens. After almost 2.5 years of effort in this line of research, we have finally succeeded in making our robot hand "spin pens."

Excited to share a new humanoid robot platform we’ve been working on. Berkeley Humanoid is a reliable and low-cost mid-scale research platform for learning-based control. We demonstrate the robot walking on various terrains and dynamic hopping with a simple RL controller.

🤖 Introducing UniT, a tactile representation learning framework that 🧠 Simple-to-Train: Trained with data from a single simple object. 🥂 Plug-and-Play: Learns a unified representation adaptable to various tasks and objects. See thread for more details (1/4)

Silicon Valley can keep convincing themselves they are the best at “software”, “AI”, until a Chinese hardware startup pulls off the best AI-driven locomotion on a production humanoid. New hardware companies from China won’t be satisfied with manufacturing, unlike their

Amazing that you can now train ACT on a macbook, with built-in camera for observation and a <$500 robot to run the policy. A large step in democratizing robot learning. Congrats Remi Cadene Hugging Face!

"We design and manufacture low-cost and easy-to-use devices for collecting human demonstration data" 🤔 they forgot to mention git clone umi-gripper.github.io