Motion-Blender Gaussian Splatting for Dynamic Scene Reconstruction
Xinyu Zhang, Haonan Chang, Yuhan Liu, Abdeslam Boularias · 2025
Read Motion-Blender Gaussian Splatting for Dynamic Scene Reconstruction by Xinyu Zhang, Haonan Chang, Yuhan Liu, Abdeslam Boularias (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-09040.txt.
Description
Xinyu Zhang, Haonan Chang, Yuhan Liu, and Abdeslam Boularias rebuild moving scenes in three dimensions so the motion itself can be edited, not just replayed. Gaussian splatting renders scenes from clouds of soft blobs, but rival methods hide motion inside neural networks or per-blob numbers, leaving nothing to grab and change. Writing in twenty twenty-five, the team represents motion as a sparse graph of links, like a skeleton. Movement of the links is blended onto nearby blobs through skinning weights that the system paints and learns, and graph and blobs are fitted together from ordinary video.
On a demanding phone video data set the method led the field, and it stayed competitive on a standard benchmark. Because motion is explicit, the team could pose objects in new ways, synthesize robot demonstrations, and plan robot actions visually. Code and demos are posted.
Catalog record
| CREATOR | Xinyu Zhang, Haonan Chang, Yuhan Liu, Abdeslam Boularias |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-09040