Universal Humanoid Robot Pose Learning from Internet Human Videos

Jiageng Mao*, Siheng Zhao*, Siqi Song*, Chuye Hong, Tianheng Shi, Junjie Ye, Mingtong Zhang, Haoran Geng, Jitendra Malik, Vitor Guizilini, Yue Wang (* equal contribution)

International Conference on Humanoid Robots (Humanoids) 2025

We propose the first scalable approach for learning generalizable and universal pose control for humanoid robots using Internet videos, which includes Humanoid-X, a massive dataset of 20M+ robot poses with text descriptions, and UH-1, a language-driven motion generation model.

[Project] [Paper] [Code]