cyberorigin/cyber_pipette

---
license: mit
---

## CyberOrigin Dataset

Our data includes information from home services, the logistics industry, and laboratory scenarios.
For more details, please refer to our [Offical Data Website](https://cyberorigin2077.github.io/)

contents of the dataset:
```
cyber_pipette # dataset root path
└── data/
├── metadata_ID1_240808.json
├── segment_ids_ID1_240808.bin # for each frame segment_ids uniquely points to the segment index that frame i came from. You may want to use this to separate non-contiguous frames from different videos (transitions).
├── videos_ID1_240808.bin # 16x16 image patches at 30hz, each patch is vector-quantized into 2^18 possible integer values. These can be decoded into 256x256 RGB images using the provided magvit2.ckpt weights.
├── ...
└── ...
```

```json
{
"task": "Pipette",
"total_episodes": 8589,
"total_frames": 3352047,
"token_dtype": "uint32",
"vocab_size": 262144,
"fps": 30,
"manipulation_type": "Bi-Manual",
"language_annotation": "None",
"scene_type": "Table Top",
"data_collect_method": "Directly Collection on Human"
}
```

许可证：MIT许可证
## CyberOrigin数据集
本数据集涵盖家庭服务、物流行业及实验室场景的相关数据。
如需了解更多细节，请访问我们的[官方数据网站](https://cyberorigin2077.github.io/)

### 数据集内容

cyber_pipette # 数据集根目录
└── data/
├── metadata_ID1_240808.json
├── segment_ids_ID1_240808.bin # 针对每一帧，segment_ids（段索引）唯一指向该帧所属的片段索引，可用于区分来自不同视频的非连续帧（转场帧）。
├── videos_ID1_240808.bin # 以30Hz帧率采集的16×16图像块，每个图像块已通过向量量化为2^18种可能的整数值。可通过配套的magvit2.ckpt权重文件解码为256×256的RGB图像。
├── ...
└── ...

json
{
"任务（task）": "移液操作（Pipette）",
"总回合数": 8589,
"总帧数": 3352047,
"Token数据类型（token_dtype）": "uint32",
"词表大小": 262144,
"帧率（fps）": 30,
"操作类型（manipulation_type）": "双手操作（Bi-Manual）",
"语言标注（language_annotation）": "无",
"场景类型（scene_type）": "桌面场景（Table Top）",
"数据采集方式（data_collect_method）": "人类直接采集"
}