File size: 3,178 Bytes
e468589
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
---
license: apache-2.0
task_categories:
- robotics
size_categories:
- 1M<n<10M
tags:
- LeRobot
- so101
- teleoperation
- stacking
- pick-and-place
configs:
- config_name: default
  data_files: data/*/*.parquet
language:
- en
pretty_name: stack_white_blue_black
---

# stack_3blocks_white_blue_black

Single-arm SO-101 teleoperation dataset for a **3-block stacking** task (bottom → top: white, blue, black), recorded with **2 cameras** in [LeRobot](https://github.com/huggingface/lerobot) format (`v3.0`).

## Task

> Stack the blocks from bottom to top: white, blue, black.

Pipeline: pick white → place as base → pick blue → stack on white → pick black → stack on blue → hold stable.

Scene: three blocks start flat and spread in an ~A4 workspace (randomized each episode, gap ≥5 cm); no tray / distractors.

## Dataset summary

| Field | Value |
| --- | --- |
| Robot | `so101_follower` (single SO-101 arm) |
| Collection | Leader–follower teleoperation |
| Episodes | **199** |
| Frames | **122,138** |
| FPS | **30** |
| Episode length | ~261–1079 frames (~8.7–36.0 s), mean ~614 frames (~20.5 s) |
| Tasks | 1 |
| Split | `train`: episodes `0:199` |
| Format | LeRobot codebase `v3.0` |
| Approx. size | ~690 MB |

## Cameras

| Key | Resolution | Role |
| --- | --- | --- |
| `observation.images.front` | 640 × 480 | Overhead / front view (UGREEN) |
| `observation.images.wrist` | 640 × 480 | Wrist camera (Sonix USB2.0 CAM1) |

All videos: H.264, 30 FPS, RGB, no audio.

## Action / state

6-DoF joint positions:

- `shoulder_pan.pos`
- `shoulder_lift.pos`
- `elbow_flex.pos`
- `wrist_flex.pos`
- `wrist_roll.pos`
- `gripper.pos`

Features:

- `action` — commanded / teleop joint targets (`float32`, shape `[6]`)
- `observation.state` — measured joint positions (`float32`, shape `[6]`)

## Features

- `observation.images.front` / `wrist` (video)
- `observation.state`
- `action`
- `timestamp`, `frame_index`, `episode_index`, `index`, `task_index`

## Load with LeRobot

```python
from lerobot.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset(
    repo_id="my_pick_place/stack_3blocks_white_blue_black",
    root="/home/rxn/datasets/stack_3blocks_white_blue_black",
)

print(dataset.meta)
print(len(dataset), "frames,", dataset.num_episodes, "episodes")
```

Visualize locally:

```bash
lerobot-dataset-viz \
  --repo-id my_pick_place/stack_3blocks_white_blue_black \
  --root /home/rxn/datasets/stack_3blocks_white_blue_black \
  --episode-index 0
```

## Collection notes

- Recorded with LeRobot SO-101 leader–follower teleop (`get-data-stack_3blocks.sh`).
- Local root: `/home/rxn/datasets/stack_3blocks_white_blue_black`
- Hub / repo id: `my_pick_place/stack_3blocks_white_blue_black`
- Success criterion: bottom white, middle blue, top black, stack standing stably before ending the episode.
- One frozen junk episode (original ep70, ~3.2 s) was removed; cleaned set is 199 episodes.
- Backup with the junk episode kept at: `/home/rxn/datasets/stack_3blocks_white_blue_black_with_ep70`

## Citation

If you use this dataset, please cite [LeRobot](https://github.com/huggingface/lerobot) and this repository.