KillzoneCCO commited on
Commit
4845dfc
·
verified ·
1 Parent(s): 05fc398

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +30 -31
README.md CHANGED
@@ -2,48 +2,47 @@
2
  tags:
3
  - robotics
4
  - lerobot
 
 
 
 
5
  ---
6
 
7
- # T1
8
 
9
- <!-- HBR:GENERATED:START -->
10
- ## Export facts
11
 
12
- - Format: LeRobot v3
13
- - Collection class: `production_candidate`
14
- - Output episodes: 288
15
- - Frames / samples: 8785
16
- - Output rate: 8.000 Hz
17
- - Original source time: 2416.985 s
18
- - Selected source time: 1098.125 s
19
- - Excluded source time: 1318.860 s
20
- - Embodiment profile: `b1111111-1111-4111-8111-111111111117`
21
- - Dataset: `8b2bfd61-b76d-4cb9-929f-9c243cab2e07`
22
- - Export: `8694c8de-fadd-59f5-afbc-a478a41acd32`
23
 
24
- Named modalities, units where declared, source inputs (including wholly excluded episodes), actual intervals and selection policy are recorded in [dataset-card.json](meta/dataset-card.json). Excluded time includes policy filtering and unavailable continuous coverage. Missing units are not inferred.
 
25
 
26
- ## Limitations
27
 
28
- Training episodes are selected continuous source intervals, not independent physical recording sessions. Missing data are not reconstructed. Optional modality validity must be respected. Source Capture and HBR retain the complete recording. Quality passage is not proof of task success or suitability for a particular training objective.
 
 
 
 
 
 
 
 
 
29
 
30
- ## Open locally
31
 
32
- Use the official loader from [https://github.com/huggingface/lerobot](https://github.com/huggingface/lerobot) at commit `33cad37054c2b594ceba57463e8f11ee374fa93c`, as pinned by the DATA_NODE compatibility environment:
33
 
34
  ```python
35
- from pathlib import Path
36
  from lerobot.datasets.lerobot_dataset import LeRobotDataset
37
 
38
- dataset = LeRobotDataset(
39
- repo_id="owner/name", root=Path("/path/to/export"),
40
- download_videos=False, video_backend="pyav",
41
- )
42
- print(len(dataset), dataset[0])
43
- ```
44
 
45
- <!-- HBR:GENERATED:END -->
46
-
47
- ## Owner notes
48
-
49
- Add the task description, intended uses and limitations here. No dataset license is inferred or granted by this export.
 
 
2
  tags:
3
  - robotics
4
  - lerobot
5
+ - booster-t1
6
+ - recovery
7
+ task_categories:
8
+ - robotics
9
  ---
10
 
11
+ # Booster T1 Task Recovery Dataset (LeRobot v3.0)
12
 
13
+ ## 🎥 Video Stream Preview (Head RGB Camera @ 8.0 FPS)
 
14
 
15
+ <video controls width="720" src="https://huggingface.co/datasets/KillzoneCCO/Booster-T1-Recovery/resolve/main/videos/observation.images.head_camera/chunk-000/file-000.mp4"></video>
 
 
 
 
 
 
 
 
 
 
16
 
17
+ > ℹ️ **Прямой просмотр видеопотока:** Видеофайл эпизода 0: [file-000.mp4](https://huggingface.co/datasets/KillzoneCCO/Booster-T1-Recovery/resolve/main/videos/observation.images.head_camera/chunk-000/file-000.mp4).
18
+ > Все 288 видео доступны в каталоге `videos/observation.images.head_camera/chunk-000/` (от `file-000.mp4` до `file-287.mp4`).
19
 
20
+ ---
21
 
22
+ ## 📊 Dataset Specifications
23
+ - **Codebase Version:** LeRobot v3.0 (`meta/info.json`)
24
+ - **Robot:** Booster T1 (`Robot-bc43b7`)
25
+ - **Task:** Transferring plates from the table to the container (Automated pick-and-place with teleoperation recovery)
26
+ - **Total Episodes:** 288 continuous manipulation segments (30 parent teleoperation recoveries)
27
+ - **Total Frames:** 8,785 frames @ 8.0 FPS
28
+ - **Action Shape:** `[20]` DOF (float32, radians)
29
+ - **State Shape:** `[37]` DOF (float32, joint positions and base pose)
30
+ - **Primary Camera:** `observation.images.head_camera` (1280x720 H.264 @ 8.0 FPS)
31
+ - **QC Quality:** 100% Passed (0 NaNs, 0 Infs, 0 Timestamp Gaps)
32
 
33
+ ---
34
 
35
+ ## 🚀 How to Load in Python (Official LeRobot v3 Loader)
36
 
37
  ```python
 
38
  from lerobot.datasets.lerobot_dataset import LeRobotDataset
39
 
40
+ # Загрузка полного датасета с Hugging Face:
41
+ dataset = LeRobotDataset("KillzoneCCO/Booster-T1-Recovery")
 
 
 
 
42
 
43
+ print(f"Total episodes: {dataset.meta.total_episodes}")
44
+ print(f"Total frames: {len(dataset)}")
45
+ sample = dataset[0]
46
+ print(f"Action shape: {sample['action'].shape}")
47
+ print(f"State shape: {sample['observation.state'].shape}")
48
+ ```