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