record-red-block-v2 / README.md
Martsinian's picture
Update README.md
60e84d2 verified
---
license: apache-2.0
task_categories:
- robotics
tags:
- LeRobot
- robotics
- robot-learning
- imitation-learning
- manipulation
- pick-and-place
- so101
configs:
- config_name: default
data_files: data/*/*.parquet
---
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Summary
- **Task:** Move a **red block** into a **white box** (block-in-bin / pick-and-place).
- **Scene setup:** The **block and the box stay in the same fixed positions** across episodes.
- **Variation:** The **red block orientation (rotation) changes** between episodes, while its position remains the same.
- **Learning goal:** Provide a clean baseline dataset for imitation learning and debugging robot learning pipelines with minimal scene variation.
## Supported Tasks and Use Cases
This dataset is primarily intended for:
- **Imitation learning / behavior cloning** (learning a policy from demonstrations).
- **Policy evaluation and benchmarking** in a low-variance environment.
- **Ablations** (e.g., impact of rotation variation on grasp strategy).
- **Sanity checks** for data loading, temporal stacking/windowing, and training loops.
## Task Description
**Instruction:** “Place the red block into the white box.”
A typical episode consists of the robot starting from a resting pose, approaching the block, grasping it, moving to the box, releasing the block into the box, and optionally returning to rest.
Because the block position is constant and only its rotation varies, the task focuses on:
- grasp approach robustness to **object yaw/roll/pitch** changes (depending on how rotation was applied),
- consistent placement motion into a fixed receptacle.
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 50,
"total_frames": 23914,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:50"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.gripper": {
"dtype": "video",
"shape": [
1080,
1920,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 1080,
"video.width": 1920,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.top": {
"dtype": "video",
"shape": [
1080,
1920,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 1080,
"video.width": 1920,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
📊 Interested in the dataset? Contact Unidata to discuss purchase terms: [https://unidata.pro](https://unidata.pro/datasets/lerobot-so-101-manipulations/?utm_source=huggingface-Martsinian&utm_medium=referral&utm_campaign=record-red-block)
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
```