This egocentric video dataset captures whole-body interactions with a range of everyday objects in ordinary living environments, recorded for Physical AI model training.
Potential Use Cases
- Humanoid and Mobile Manipulation Policy Training: The recorded actions require posture, reach, and locomotion rather than hand motion alone, supplying demonstrations for embodiments that move their whole body to complete a task.
- Egocentric Activity Understanding: First-person footage in which the camera moves with the actor supports action recognition and temporal segmentation under the viewpoint shift that fixed-camera datasets never produce.