Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- robotics
|
| 5 |
+
tags:
|
| 6 |
+
- LeRobot
|
| 7 |
+
- so101
|
| 8 |
+
- vision-language-action
|
| 9 |
+
- robotics
|
| 10 |
+
- manipulation
|
| 11 |
+
dataset_info:
|
| 12 |
+
features:
|
| 13 |
+
- name: observation.state
|
| 14 |
+
sequence: float32
|
| 15 |
+
- name: action
|
| 16 |
+
sequence: float32
|
| 17 |
+
- name: timestamp
|
| 18 |
+
dtype: float32
|
| 19 |
+
- name: frame_index
|
| 20 |
+
dtype: int64
|
| 21 |
+
- name: episode_index
|
| 22 |
+
dtype: int64
|
| 23 |
+
- name: index
|
| 24 |
+
dtype: int64
|
| 25 |
+
- name: task_index
|
| 26 |
+
dtype: int64
|
| 27 |
+
- name: observation.images.top
|
| 28 |
+
dtype: video
|
| 29 |
+
- name: observation.images.wrist
|
| 30 |
+
dtype: video
|
| 31 |
+
splits:
|
| 32 |
+
- name: train
|
| 33 |
+
num_bytes: 356000
|
| 34 |
+
num_examples: 5944
|
| 35 |
+
---
|
| 36 |
+
|
| 37 |
+
# SO101 VLA Dataset v2.1
|
| 38 |
+
|
| 39 |
+
This dataset contains 20 episodes of robot demonstrations for the "pick up the black tape" task using the SO101 follower robot.
|
| 40 |
+
|
| 41 |
+
## Dataset Details
|
| 42 |
+
|
| 43 |
+
- **Format**: LeRobot v2.1
|
| 44 |
+
- **Episodes**: 20
|
| 45 |
+
- **Total Frames**: 5,944
|
| 46 |
+
- **FPS**: 30
|
| 47 |
+
- **Robot Type**: so101_follower
|
| 48 |
+
- **Task**: pick up the black tape
|
| 49 |
+
|
| 50 |
+
## Features
|
| 51 |
+
|
| 52 |
+
- **Actions**: 6D joint positions (shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos)
|
| 53 |
+
- **States**: 6D joint positions (same as actions)
|
| 54 |
+
- **Images**:
|
| 55 |
+
- Top camera: 720x1280 RGB
|
| 56 |
+
- Wrist camera: 240x320 RGB
|
| 57 |
+
- **Actions are UNNORMALIZED** (in degrees)
|
| 58 |
+
|
| 59 |
+
## Dataset Structure
|
| 60 |
+
|
| 61 |
+
```
|
| 62 |
+
data/chunk-000/episode_XXXXXX.parquet # Per-episode data files
|
| 63 |
+
videos/chunk-000/observation.images.top/episode_XXXXXX.mp4
|
| 64 |
+
videos/chunk-000/observation.images.wrist/episode_XXXXXX.mp4
|
| 65 |
+
meta/info.json
|
| 66 |
+
meta/episodes.jsonl
|
| 67 |
+
meta/tasks.jsonl
|
| 68 |
+
meta/episodes_stats.jsonl
|
| 69 |
+
```
|
| 70 |
+
|
| 71 |
+
## Usage
|
| 72 |
+
|
| 73 |
+
```python
|
| 74 |
+
from lerobot.datasets.lerobot_dataset import LeRobotDataset
|
| 75 |
+
|
| 76 |
+
dataset = LeRobotDataset("abuod0/so101_vla_dataset_v21")
|
| 77 |
+
```
|
| 78 |
+
|
| 79 |
+
## Citation
|
| 80 |
+
|
| 81 |
+
```bibtex
|
| 82 |
+
@dataset{so101_vla_dataset_v21,
|
| 83 |
+
title={SO101 VLA Dataset v2.1},
|
| 84 |
+
author={Your Name},
|
| 85 |
+
year={2024},
|
| 86 |
+
url={https://huggingface.co/datasets/abuod0/so101_vla_dataset_v21}
|
| 87 |
+
}
|
| 88 |
+
```
|
| 89 |
+
|