✨New #IROS2024 paper: “Adapting Skills to Novel Grasps: A Self-Supervised Approach”
We leverage self-supervised data to adapt skill trajectories to novel object grasp poses.
No need for depth, no prior knowledge, no CAD models, no calibration - just RGB images.
Here's how👇
When your work ends up all the way to the cover of @SciRobotics you can’t help it but feel incredibly honoured and proud.
It’s been an absolute pleasure working with my co-lead and dear friend @kamildre. And a great thanks goes to the entire team @vitalisvos19@Ed__Johns . Mega job, well done.
Special thanks also go to the @SciRobotics reviewers and editors who helped improve the paper and made the submission process much more enjoyable.
Excited to share Gemini Robotics On-Device, a VLA that runs locally on a single GPU. ⚡️
It combines dexterity, robustness, and inference speed, and can be finetuned easily to solve new tasks and on new embodiments.
If you're at RSS, there is going to be a live session today!
A great chat about robots that learn by watching you, in-context imitation learning, how to use foundation models and more.
Watch it here 👇🏻
Thanks @chris_j_paxton and @micoolcho for having us and to my friend and collaborator @geopgs for the long nights spent on this project.
Imagine a robot that learns all sorts of tasks simply by watching you.
This future is closer than ever: R+X was accepted at ICRA 2025.
From an hour long POV video we can perform in-context imitation learning, from human to robot. No training or robot data needed.
Posting a call for help: does anyone know of a good way to simultaneously treat both POTS and Ménière’s disease? Please contact me if you’re either a clinician with experience doing this or a patient who has found a good solution. Context in thread
@DotanDiCastro@Ed__Johns@DotanDiCastro Thanks for sharing InsertionNet. The goal of MILES was to develop a general imitation learning method from a single human demo, that does precise tasks, including insertion, but also more general tasks like opening a lid or placing a bread into a toaster.
Want to teach your robot new tasks from only a single demo?
We've just released code for MILES, which we presented at CoRL 2024 last week.
Learning is fully automated: you just provide a single demonstration, then sit back and relax! 🍹😴
Code: https://t.co/UD2ZvKtlq0.
🧵👇
Could robots collect and self-label their own data? It would be much easier than manually giving endless demos!
New at #CoRL2024:
🦾✌️ MILES: Making Imitation Learning Easy with Self-Supervision
Our robot learned to use a key from just one demo!
https://t.co/UD2ZvKtlq0
More 👇
Can a robot learn a skill with one object grasp, and then perform that skill with an entirely different object grasp?
Today at #IROS2024, @geopgs will present our solution based on self-supervised learning! 🤖📷🔨
Oral at 6pm in Room 3. Come along 😀.
https://t.co/FH0nfm9WCi
This work could not have been possible without the great help of my coauthors @vitalisvos19 and Kamil Dreczkowski in implementing the baseline methods and useful discussions.
✨New #IROS2024 paper: “Adapting Skills to Novel Grasps: A Self-Supervised Approach”
We leverage self-supervised data to adapt skill trajectories to novel object grasp poses.
No need for depth, no prior knowledge, no CAD models, no calibration - just RGB images.
Here's how👇
For more insights and additional experiment results, see our arxiv paper and website:
Paper: https://t.co/d6bCo7ZXDs
Website: https://t.co/J6u6XpgkCy
Supervised by @Ed__Johns