---
license: apache-2.0
---
This is a multi-class version of open vocabulary segmentation dataset by randomly merging annotations from several classes, including ADE20K(A-150), PASCAL Context59(PC-59), and PASCAL VOC20(PAS-20).
You can use them according to [\[📂 GitHub\]](https://github.com/yayafengzi/ALToLLM).

If you find this project useful in your research, please consider citing:

```BibTeX
@article{wang2025alto,
  title={ALTo: Adaptive-Length Tokenizer for Autoregressive Mask Generation},
  author={Wang, Lingfeng and Lin, Hualing and Chen, Senda and Wang, Tao and Cheng, Changxu and Zhong, Yangyang and Zheng, Dong and Zhao, Wuyue},
  journal={arXiv preprint arXiv:2505.16495},
  year={2025}
}
```