Phantom: Training Robots Without Robots Using Only Human Videos
Marion Lepert 1 , Jiaying Fang 1 , Jeannette Bohg 1 · 2025
Read Phantom: Training Robots Without Robots Using Only Human Videos by Marion Lepert 1 , Jiaying Fang 1 , Jeannette Bohg 1 (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-2503-00779.txt.
Description
Marion Lepert, Jiaying Fang, and Jeannette Bohg train robot arms from human videos alone, in a 2025 paper. Teaching robots usually needs costly sessions where a person drives the robot itself. Human video, by contrast, is everywhere. Their framework, Phantom, turns such videos into robot-ready lessons.
Hand tracking extracts the motion, then the human arm is painted out and a rendered robot arm drawn in. The edited clips become matching sight and action pairs for training. Policies learned this way ran on real hardware with no extra tuning. Success reached up to ninety-two percent on tasks from folding soft objects to sweeping items and inserting parts. The approach could let robots learn from the vast stock of human demonstration online.
Catalog record
| CREATOR | Marion Lepert 1 , Jiaying Fang 1 , Jeannette Bohg 1 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-00779