Egostation Gopro Pick And Place V1: First-Person Robotic Task Demonstrations

First-person human demonstration data for a donut pick-and-place task. The data was captured with a head-mounted GoPro Hero 9/11 camera and processed by the ZenO Studio pipeline. It is formatted for the LeRobot v2.1 training pipeline.

Use Cases

• Train robotic imitation learning models based on first-person video demonstrations.
• Benchmark visual perception algorithms for object manipulation tasks based on the described pick-and-place scenario.
• Develop and test state estimation models using the described time series and hand keypoint data.
• Validate robotic control policies in simulation using real-world demonstration trajectories.

Strengths

• Data is captured in 4K HEVC video format, suggesting high visual fidelity.
• The dataset is pre-processed and formatted to drop directly into the LeRobot training pipeline, reducing setup effort.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.

Provenance

Source

zeno-labs

Collection Method

Captured via head-mounted GoPro Hero 9/11 camera and processed by the ZenO Studio pipeline.

Freshness

Last updated 2026-05-12 21:18:39; freshness should be verified.