| { |
| "schema_version": 1, |
| "episodes": 134537, |
| "frames": 42559113, |
| "shards": 268, |
| "bytes": 4257792000, |
| "counts": { |
| "episodes": 134537, |
| "frames": 42559113, |
| "copied_command_episodes": 0, |
| "unresolved_sparse_command_episodes": 0, |
| "command_review_episodes": 12, |
| "state_invalid_arm_frames": 0, |
| "command_invalid_arm_frames": 3654 |
| }, |
| "coverage_complete": true, |
| "all_shards_checksum_verified": true, |
| "all_frames_pose_schema_verified": true, |
| "data_index": "eef/episodes.jsonl", |
| "shard_manifest": "eef/shards.jsonl", |
| "statistics": "eef/statistics.json", |
| "reader": "eef/reader.py", |
| "pose_format": "float32 XYZ metres + xyzw quaternion, [frames,2,7]", |
| "actions": "Recorded-command poses remain separate from explicitly derived next-state targets. Select action_mode in the reader.", |
| "camera_reprojection_validation": "Not certified across the dataset. RoboMIND poses are nominal-model FK, with explicit per-arm base and terminal-arm frame names.", |
| "models": "eef/models.json" |
| } |
|
|
