Datasets:
| pretty_name: Human Ego Data | |
| tags: | |
| - robotics | |
| - egocentric-video | |
| - rgb-d | |
| - human-demonstrations | |
| # Human Ego Data | |
| Example RGB-D episodes collected with the | |
| [ego_realsense_tool](https://github.com/baiyu858/ego_realsense_tool) for the | |
| Ego2Sim benchmark. | |
| Only tasks with completed successful recordings on the collection machine are | |
| included. Each currently available task contributes one raw example episode. | |
| ## Layout | |
| ```text | |
| Txx_task_name/ | |
| suite/ | |
| session_xxx/ | |
| raw/ | |
| frames/ | |
| *_color.png | |
| *_depth.png | |
| frames.csv | |
| imu.csv | |
| intrinsics.json | |
| session.json | |
| capture_stats.json | |
| control_events.jsonl | |
| ``` | |
| Depth images are 16-bit PNG files aligned to the RGB camera. Camera intrinsics | |
| and the depth scale are stored in `intrinsics.json`. Episode instructions, | |
| task/suite identities, capture settings, and completion state are stored in | |
| `session.json`. | |
| ## Included Examples | |
| | Task | Suite | Episode | Frames | Instruction | | |
| | --- | --- | --- | ---: | --- | | |
| | `T01_pick_place_block` | `object` | `session_000` | 121 | `[right+white tablecloth]Pick up the green block and place it in the green plate.` | | |
| | `T04_put_object_cabinet` | `spatial` | `session_000` | 315 | `[left+black]Open the bottom drawer and put the red block inside it.` | | |
| See `manifest.json` for machine-readable sizes and paths. | |