Twitxy Saved

Carl Doersch

@CarlDoersch

Researcher at DeepMind
London, UK carldoersch.com Joined April 2017
290 Following 2.5K Followers 67 Posts

Latest posts 5

Carl Doersch @CarlDoersch
Huge improvements in TAPNext's ability to track through long videos, now available for download! This was an amazing collaboration with Google's Android XR team, thanks @sjung0 for all the hard work!
馃殌 Exciting News: TAPNext++ will be presented at CVPR 2026!

I鈥檓 thrilled to share our latest work, "TAPNext++: What's Next for Tracking Any Point (TAP)?", completed during my time as a Student Researcher at Google.
馃У馃憞
View quoted post on X
0 replies 4 reposts 22 likes 2.8K views View on X
Carl Doersch @CarlDoersch
I "found" a few of the tasks in this video, and it's hard to convey the feeling to those who don't know the training data. Just know that a few episodes started with someone saying something like "this robot hasn't seen anything like this task; I doubt it'll work..."
Konstantinos Bousmalis Verified account @bousmalis Oct 1, 2025
You have to watch this! For years now, I've been looking for signs of nontrivial zero-shot transfer across seen embodiments. When I saw the Alohas unhang tools from a wall used only on our Frankas I knew we had it! Gemini Robotics 1.5 is the first VLA to achieve such transfer!!
View quoted post on X
1 replies 3 reposts 9 likes 1.2K views View on X
Carl Doersch @CarlDoersch
Want a robot to solve a task, specified in language? Generate a video of a person doing it, and then retarget the action to the robot with the help of point tracking! Cool collab with @mangahomanga during his student researcher stint at Google.
Homanga Bharadhwaj @ CoRL Verified account @mangahomanga Sep 24, 2024
Gen2Act: Casting language-conditioned manipulation as *human video generation* followed by *closed-loop policy execution conditioned on the generated video* enables solving diverse real-world tasks unseen in the robot dataset!

homangab.github.io/gen2act/

1/n
View quoted post on X