Doc-750K

This dataset is in the paper [Docopilot: Improving Multimodal Models for Document-Level Understanding](https://arxiv.org/abs/2507.14675).

Please refer to https://github.com/OpenGVLab/Docopilot for details.

## FAQ
### Unzipping Split Archives on Linux
If you encounter issues when unzipping the image archive on Linux, such as:
- zip bomb warnings
- bad zipfile offset errors

Please try the following solutions:

1. Zip Bomb Warning
Some systems may trigger a zip bomb detection warning due to the large number of small image files. You can bypass this by disabling the detection with:
```bash
export UNZIP_DISABLE_ZIPBOMB_DETECTION=TRUE
```
2. Bad Zipfile Offset Error
If you're dealing with split zip archives (e.g., images.z01, images.z02, ..., images.zip), you need to merge them before unzipping:

```bash
zip -s 0 images.zip --out images_full.zip
unzip images_full.zip
```
This will reconstruct the full archive and allow you to unzip it normally.

Note: The image dataset is very large, so please ensure you have sufficient disk space and patience during extraction.

本数据集来自论文《Docopilot：面向文档级理解的多模态模型优化》（https://arxiv.org/abs/2507.14675）。
详细信息请参考https://github.com/OpenGVLab/Docopilot。

### Linux系统下拆分压缩包的解压问题
若您在Linux系统中解压图像压缩包时遇到以下问题：
- 压缩炸弹警告
- ZIP文件偏移错误

请尝试以下解决方案：

1. 压缩炸弹警告
部分系统会因包含大量小型图像文件而触发压缩炸弹检测警告，可通过以下命令关闭检测以绕过该问题：
bash
export UNZIP_DISABLE_ZIPBOMB_DETECTION=TRUE

2. ZIP文件偏移错误
若您遇到的是拆分式ZIP压缩包（例如images.z01、images.z02……images.zip），则需先合并再解压：
bash
zip -s 0 images.zip --out images_full.zip
unzip images_full.zip

该命令可重建完整压缩包，随后即可正常解压。

注意：本图像数据集体积较大，请确保拥有足够的磁盘空间，并在解压过程中保持耐心。