API for multimodal
human activity data

Egocentric human activity data with SLAM, 3D pose, hand reconstruction, and human labels.

Built by a team from

Amazon
Stanford University
University of Michigan
National University of Singapore

The largest multimodal dataset of human activity

Egocentric video, motion, and sensor data collected from real workers across dozens of industries worldwide.

#Industry
1Manufacturing
2Construction
3Restaurants
4Hospitality
5Electrical Work
6Warehousing
7Agriculture
8Automotive
9Healthcare
10Household

Data modalities

Egocentric video

First-person footage from real workers across real environments

Motion & sensor data

IMU, accelerometer, and gyroscope streams from wearable devices

Dense annotations

Dense, human annotations on every video. Included with each dataset or available standalone.

Build it with us

Renlei works with people and workplaces around the world to capture real activity responsibly, with clear instructions and scope agreed in advance. Choose one or more ways to get involved.

Contribute to projects

Record approved tasks with wearable equipment on paid, clearly scoped projects.

Make introductions

Connect Renlei with people, communities, or companies that could support local projects.

Lead in your region

Recruit, coordinate, and support contributors and projects in your city or country.

Host a project

Explore approved data collection inside a factory, farm, warehouse, workshop, or site.