---
title: Kai0
canonical_url: "https://www.modelscope.cn/datasets/OpenDriveLab/Kai0"
md_url: "https://www.modelscope.cn/datasets/OpenDriveLab/Kai0.md"
repository: OpenDriveLab/Kai0
last_updated: 2026-06-27
license: cc-by-nc-sa-4.0
storage_size: "360 GB"
downloads: 1586232
stars: 1
---
# Kai0
> Kai0 - OpenDriveLab 在 ModelScope 开源的数据集。TODO [ ] The advantage label will be coming soon.
OpenDriveLab/Kai0 是 ModelScope 魔搭社区上的数据集，存储大小 360 GB，采用 cc-by-nc-sa-4.0 许可。
- **Repository**: OpenDriveLab/Kai0
- **License**: cc-by-nc-sa-4.0
- **Storage size**: 360 GB
- **Downloads**: 1586232
- **Stars**: 1
- **Last updated**: 2026-06-27
Source: https://www.modelscope.cn/datasets/OpenDriveLab/Kai0
---
# KAI0
# TODO
- [ ] The advantage label will be coming soon.
## Contents
- [About the Dataset](#about-the-dataset)
- [Load the Dataset](#get-started)
- [Download the Dataset](#download-the-dataset)
- [Dataset Structure](#dataset-structure)
- [Folder hierarchy](#folder-hierarchy)
- [Details](#details)
- [License and Citation](#license-and-citation)
## [About the Dataset](#contents)
- **~134 hours** real world scenarios
- **Main Tasks**
- ***Task_A***
- Single task
- Initial state: T-shirts are randomly tossed onto the table, presenting random crumpled configurations
- Manipulation task: Operate the robotic arm to unfold the garment, then fold it
- ***Task_B***
- Garment classification and arrangement task
- Initial state: Randomly pick a garment from the laundry basket
- Classification: Determine whether the garment is a T-shirt or a dress shirt
- Manipulation task:
- If it is a T-shirt, fold the garment
- If it is a dress shirt, expose the collar, then push it to one side of the table
- ***Task_C***
- Single task
- Initial state: Hanger is randomly placed, garment is randomly positioned on the table
- Manipulation task: Operate the robotic arm to thread the hanger through the garment, then hang it on the rod
- **Count of the dataset**
| Task | Base (episodes count/hours) | DAgger (episodes count/hours) | Total(episodes count/hours) |
|------|-----------------------------|-------------------------------|-----------------------------|
| Task_A | 3,055/~42 hours | 3,457/ ~13 Hours | 6,512 /~55 hours |
| Task_B | 5988/~31 hours | 769/~22 hours | 6757/~53 hours |
| Task_C | 6954/~61 hours | 686/~12 hours | 7640/~73 hours |
| **Total** | **15,997/~134 hours** | **4,912/~47 hours** | **20,909/~181 hours** |
## [Load the dataset](#contents)
- This dataset was created using [LeRobot](https://github.com/huggingface/lerobot)
- The dataset's version is LeRobotDataset v2.1
### For LeRobot version < 0.4.0
Choose the appropriate import based on your version:
| Version | Import Path |
|------------------------|-------------|
| `<= 0.1.0` | `from lerobot.common.datasets.lerobot_dataset import LeRobotDataset` |
| `> 0.1.0` and `< 0.4.0` | `from lerobot.datasets.lerobot_dataset import LeRobotDataset` |
```python
# For version <= 0.1.0
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
# For version > 0.1.0 and < 0.4.0
from lerobot.datasets.lerobot_dataset import LeRobotDataset
# Load the dataset
dataset = LeRobotDataset(repo_id='where/the/dataset/you/stored')
```
### For LeRobot version >= 0.4.0
You need to migrate the dataset from v2.1 to v3.0 first. See the official documentation: [Migrate the dataset from v2.1 to v3.0](https://huggingface.co/docs/lerobot/lerobot-dataset-v3)
```bash
python -m lerobot.datasets.v30.convert_dataset_v21_to_v30 --repo-id=
```
## [Download the Dataset](#contents)
### Git 下载
```bash
#Git模型下载
git clone https://www.modelscope.cn/datasets/OpenDriveLab/Kai0.git
```
### Terminal (CLI)
```bash
# Download a single file
modelscope download --dataset OpenDriveLab-org/kai0 README.md \
--local_dir "/where/you/want/to/save"
# Download the entire dataset
modelscope download --dataset OpenDriveLab-org/kai0 \
--local-dir "/where/you/want/to/save"
```
## [Dataset Structure](#contents)
### [Folder hierarchy](#contents)
Under each task directory, data is partitioned into two subsets: base and dagger.
- base
contains
original demonstration trajectories of robotic arm manipulation for garment arrangement tasks.
- dagger
contains on-policy recovery trajectories collected via iterative DAgger, designed to populate failure recovery modes absent in static demonstrations.
```text
Kai0-data/
├── Task_A/
│ ├── base/
│ │ ├── data/
│ │ │ ├── chunk-000/
│ │ │ │ ├── episode_000000.parquet
│ │ │ │ ├── episode_000001.parquet
│ │ │ │ └── ...
│ │ │ └── ...
│ │ ├── videos/
│ │ │ ├── chunk-000/
│ │ │ │ ├── observation.images.hand_left/
│ │ │ │ │ ├── episode_000000.mp4
│ │ │ │ │ ├── episode_000001.mp4
│ │ │ │ │ └── ...
│ │ │ │ ├── observation.images.hand_right/
│ │ │ │ │ ├── episode_000000.mp4
│ │ │ │ │ ├── episode_000001.mp4
│ │ │ │ │ └── ...
│ │ │ │ ├── observation.images.top_head/
│ │ │ │ │ ├── episode_000000.mp4
│ │ │ │ │ ├── episode_000001.mp4
│ │ │ │ │ └── ...
│ │ │ │ └── ...
│ │ │ └── ...
│ │ └── meta/
│ │ ├── info.json
│ │ ├── episodes.jsonl
│ │ ├── tasks.jsonl
│ │ └── episodes_stats.jsonl
│ └── dagger/
├── Task_B/
│ ├── base/
│ └── dagger/
├── Task_C/
│ ├── base/
│ └── dagger/
└── README.md
```
### [Details](#contents)
#### info.json
the basic struct of the [info.json](#meta/info.json)
```json
{
"codebase_version": "v2.1",
"robot_type": "agilex",
"total_episodes": ..., # the total episodes in the dataset
"total_frames": ..., # The total number of video frames in any single camera perspective
"total_tasks": ..., # Total number of tasks
"total_videos": ..., # The total number of videos from all camera perspectives in the dataset
"total_chunks": ..., # The number of chunks in the dataset
"chunks_size": ..., # The max number of episodes in a chunk
"fps": ..., # Video frame rate per second
"splits": { # how to split the dataset
"train": ...
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"observation.images.top_head": { # the camera perspective
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channel"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.hand_left": { # the camera perspective
...
},
"observation.images.hand_right": { # the camera perspective
...
},
"observation.state": {
"dtype": "float32",
"shape": [
14
],
"names": null
},
"action": {
"dtype": "float32",
"shape": [
14
],
"names": null
},
"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
}
}
}
```
#### [Parquet file format](#contents)
| Field Name | shape | Meaning |
|------------|-------------|-------------|
| observation.state | [N, 14] |left `[:, :6]`, right `[:, 7:13]`, joint angle
left`[:, 6]`, right `[:, 13]` , gripper open range|
| action | [N, 14] |left `[:, :6]`, right `[:, 7:13]`, joint angle
left`[:, 6]`, right `[:, 13]` , gripper open range |
| timestamp | [N, 1] | Time elapsed since the start of the episode (in seconds) |
| frame_index | [N, 1] | Index of this frame within the current episode (0-indexed) |
| episode_index | [N, 1] | Index of the episode this frame belongs to |
| index | [N, 1] | Global unique index across all frames in the dataset |
| task_index | [N, 1] | Index identifying the task type being performed |
### [tasks.jsonl](#Task_A/meta/tasks.jsonl)
Contains task language prompts (natural language instructions) that specify the manipulation task to be performed. Each entry maps a task_index to its corresponding task description, which can be used for language-conditioned policy training.
# License and Citation
All the data and code within this repo are under [](). Please consider citing our project if it helps your research.
```BibTeX
@misc{,
title={},
author={},
howpublished={\url{}},
year={}
}
