Hongyuan Chen

[Anpei Chen @AnpeiC]

1/5 Excited to share an early exploration.

Coding agents like Codex can already tackle much of real-to-sim—even for robot manipulation. Does that mean 3D vision is solved?

Not quite. 3D still matters.

https://t.co/1CPFCMFuhZ

[Anpei Chen @AnpeiC]

#Motion324: 3D Motion Reconstruction for 4D Synthesis

We offer a feed-forward framework that synthesizes high-quality 4D assets from just a single monocular video.

 ✅ Mesh
 ✅ 3D Motion
 ✅ Feed-Forward
 ✅ Motion Retargeting

Check out 👇
https://t.co/b6S5aazoJD

[Xiao-Xiao Long @xxlong0]

Very excited to share our recent work CraftsMan3D. CraftsMan3D (aka 匠心) is a two-stage 3D native mesh diffusion model with a normal-based geometry refiner by mimicking the modeling workflow of artist/craftsman. @weiyuli99072112

The first stage is a large 3D native diffusion model which directly produces high-quality 3D meshes in seconds.
Unlike multi-view reconstruction methods such as Wonder3D or LRM, CraftsMan3D is directly trained on 3D representations and therefore produces regular surfaces and complex structures.
The best thing is that we have already released both training and inference codes of our large 3D native diffusion model. Feel free to have a try on the huggingface demo.

Project page: https://t.co/TyZmsI6pxH
Github code repo: https://t.co/X1OPlJNhrg
Huggingface demo: https://t.co/bqBp5IV18R

#3DGeneration #Diffusion #Huggingface #CraftsMan3D