iiinova — where your digital self keeps evolving

ddel

@ddonatien.bsky.social

55 followers 148 following 8 posts

Robotics researcher 🔗 ddonatien.github.io

Posts Replies Media

ddel @ddonatien.bsky.social · 30/01/2026

I have juste released a pre-print 🎉🎉 🔗 ddonatien.github.io/mopt-website/

ddonatien.github.io

Disentangling perception and reasoning for improving data efficiency in learning cloth manipulation without demonstrations

000

Reposted by ddel

Michael J. Black @michael-j-black.bsky.social · 06/02/2025

Turning videos into 3D humans has never been easier or more accurate. MoCapade 3.0 improves everything over 2.0 while adding 3D camera tracking (and output) and multi-person capture. Try out the new state-of-the-art. Of course, there's more to come.

0346

ddel @ddonatien.bsky.social · 27/01/2025

Surely not* A good point could be made here about regulating the random dopamine providers that pop up everywhere (from Las Vegas to YouTube, Insta and TikTok). But why follow the route of stigmatisation based on a few anecdotal examples?

000

ddel @ddonatien.bsky.social · 27/01/2025

I don't know the context of writing this, but it seems very shallow and misguided. Is there any actual data behind this, or is it just rambling? Sure, some people are dumb and gamble, but they are surely the majority. Is it really the prospect of money that drives gambling?

110

ddel @ddonatien.bsky.social · 11/12/2024

In Korea, traffic lights stop the whole intersection at once, whereas in France it's only a few lanes.

010

ddel @ddonatien.bsky.social · 06/12/2024

I just shared my code for my latest paper, "GARField: Addressing the visual Sim-to-Real gap in garment manipulation with mesh-attached radiance fields" Meet me Wed 11th, 10:30 at #ROBIO2024 for the oral presentation! 🔗-> ddonatien.github.io/garfield-web... #robotics #garment_manipulation

ddonatien.github.io

000

ddel @ddonatien.bsky.social · 05/12/2024

I guess they could also be classified as photogrammetry technology as it approximates 3D geometry from 2D images.

000

ddel @ddonatien.bsky.social · 05/12/2024

Super exciting work, Pr. Gupta. Regarding the real2sim approach, the paper states, "Digital twins are obtained directly from real-world videos [..] using photogrammetry methods such as Gaussian splatting and neural radiance fields". Do you use both? Only these 2? Why not settle for only one?

100

ddel @ddonatien.bsky.social · 03/12/2024

Very insightful paper, thanks. In the same vein, I found this paper (I'm not related to the authors) an interesting read: arxiv.org/abs/2310.04411

arxiv.org

Understanding, Predicting and Better Resolving Q-Value Divergence in Offline-RL

The divergence of the Q-value estimation has been a prominent issue in offline RL, where the agent has no access to real dynamics. Traditional beliefs attribute this instability to querying out-of-dis...

141