All guides

Data annotation adds labels to images, video, LiDAR, and robot episodes so models can learn. How teams design schemas, run QA, and automate labeling.

LeRobot is Hugging Face's open source library for robot learning. Learn how the LeRobotDataset v3.0 format works and how to curate data and train policies.

MCAP is the open container format for timestamped robotics data and the ROS 2 default. Learn how MCAP files work, how to open them, and how to curate them.

Multimodal data combines camera, LiDAR, IMU, and robot state streams from one moment. How robotics and physical AI teams store, curate, and label it.

Physical AI is AI that perceives and acts in the real world through robots and vehicles. Learn how it works, the models behind it, and why data decides it.