Phantom: Training Robots Without Robots Using Only Human Videos

Marion Lepert, Jiaying Fang, Jeannette Bohg
Proceedings of The 9th Conference on Robot Learning, PMLR 305:4545-4565, 2025.

Abstract

Training general-purpose robots requires learning from large and diverse data sources. Current approaches rely heavily on teleoperated demonstrations which are difficult to scale. We present a scalable framework for training manipulation policies directly from human video demonstrations, requiring no robot data. Our method converts human demonstrations into robot-compatible observation-action pairs using hand pose estimation and visual data editing. We inpaint the human arm and overlay a rendered robot to align the visual domains. This enables zero-shot deployment on real hardware without any fine-tuning. We demonstrate strong success rates—up to 92%—on a range of tasks including deformable object manipulation, multi-object sweeping, and insertion. Our approach generalizes to novel environments and supports closed-loop execution. By demonstrating that effective policies can be trained using only human videos, our method broadens the path to scalable robot learning. Videos are available at https://phantom-training-robots.github.io.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-lepert25a, title = {Phantom: Training Robots Without Robots Using Only Human Videos}, author = {Lepert, Marion and Fang, Jiaying and Bohg, Jeannette}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {4545--4565}, year = {2025}, editor = {Lim, Joseph and Song, Shuran and Park, Hae-Won}, volume = {305}, series = {Proceedings of Machine Learning Research}, month = {27--30 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v305/main/assets/lepert25a/lepert25a.pdf}, url = {https://proceedings.mlr.press/v305/lepert25a.html}, abstract = {Training general-purpose robots requires learning from large and diverse data sources. Current approaches rely heavily on teleoperated demonstrations which are difficult to scale. We present a scalable framework for training manipulation policies directly from human video demonstrations, requiring no robot data. Our method converts human demonstrations into robot-compatible observation-action pairs using hand pose estimation and visual data editing. We inpaint the human arm and overlay a rendered robot to align the visual domains. This enables zero-shot deployment on real hardware without any fine-tuning. We demonstrate strong success rates—up to 92%—on a range of tasks including deformable object manipulation, multi-object sweeping, and insertion. Our approach generalizes to novel environments and supports closed-loop execution. By demonstrating that effective policies can be trained using only human videos, our method broadens the path to scalable robot learning. Videos are available at https://phantom-training-robots.github.io.} }
Endnote
%0 Conference Paper %T Phantom: Training Robots Without Robots Using Only Human Videos %A Marion Lepert %A Jiaying Fang %A Jeannette Bohg %B Proceedings of The 9th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Joseph Lim %E Shuran Song %E Hae-Won Park %F pmlr-v305-lepert25a %I PMLR %P 4545--4565 %U https://proceedings.mlr.press/v305/lepert25a.html %V 305 %X Training general-purpose robots requires learning from large and diverse data sources. Current approaches rely heavily on teleoperated demonstrations which are difficult to scale. We present a scalable framework for training manipulation policies directly from human video demonstrations, requiring no robot data. Our method converts human demonstrations into robot-compatible observation-action pairs using hand pose estimation and visual data editing. We inpaint the human arm and overlay a rendered robot to align the visual domains. This enables zero-shot deployment on real hardware without any fine-tuning. We demonstrate strong success rates—up to 92%—on a range of tasks including deformable object manipulation, multi-object sweeping, and insertion. Our approach generalizes to novel environments and supports closed-loop execution. By demonstrating that effective policies can be trained using only human videos, our method broadens the path to scalable robot learning. Videos are available at https://phantom-training-robots.github.io.
APA
Lepert, M., Fang, J. & Bohg, J.. (2025). Phantom: Training Robots Without Robots Using Only Human Videos. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:4545-4565 Available from https://proceedings.mlr.press/v305/lepert25a.html.

Related Material