Dataset · InternRobotics
2000+ Hours of Annotated Human Motion Capture
InternRobotics/MotionMillion
A large-scale human motion dataset extracted from web video with detailed descriptions for robotics and animation generation.
What's inside
The dataset contains between 1M and 10M records (over 2,000 hours) of human motion extracted from online videos, including fitness, martial arts, dancing, performances, and sports. Each motion is paired with over 20 detailed English annotations describing age, body characteristics, movement style, emotions, and environment.
What you can do with it
- Training 3D animation and motion generation models from text descriptions (text-to-motion)
- Training control systems and locomotion for humanoid robots
- Training models for recognizing complex actions, dances, and martial arts
- Tasks
- Other, Robotics
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 1.2K
- Likes
- 58
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("InternRobotics/MotionMillion")