EgoExplore: Large-Scale Open World Exploration Dataset
6 days ago
- #egocentric-video
- #AI-research
- #dataset
- EgoExplore is a real-world dataset of egocentric video sequences for open-world exploration.
- The dataset emphasizes embodied physics, object interaction, and multi-agent dynamics.
- Collected using head-mounted cameras in natural, unstructured environments.
- Features freeform exploration with rich, continuous trajectories.
- Rich in object interaction, human action, and behavior of other agents.
- Supports research in physics-aware world models, temporal and spatial continuity, object affordances, and multi-agent dynamics.
- Dataset statistics include 62 hours of footage, 1080p to 4K resolution, and 30-60fps.
- Locations span multiple countries and diverse environments.
- Includes detailed annotations: action labels, summaries, and IMU data.
- Contact [email protected] or [email protected] for full sample dataset requests.