Fidelity-IMU-1h
Type: Egocentric, sensor-rich manipulation dataset
Duration: ~1 hour
Description:
This dataset contains everyday manipulation trajectories (doing dishes) enriched with inferred egocentric inertial signals (accelerometer + gyroscope). All data is derived from monocular video using our egocentric enrichment pipeline.
Intended use:
Robotics research, egocentric learning, embodied AI experiments.
Data format:
- Images / frames
- Camera poses
- Human joints
- Inferred IMU signals
- Object states and kinematics
Source:
Monocular video processed by Fidelity’s egocentric enrichment pipeline.
License: Apache-2.0
