Hao Zhang, Fang Li, Lu Qi, Ming-Hsuan Yang, Narendra Ahuja

[arXiv AAAI]

• A plug-in module for OOD, ZS3 and DA semantic segmentation.
• Rank 1 @ SMIYC-Anomaly Track & Obstacle Track (w/o OOD) on mean F1

• 2024.1.17 Add raw code.

See installation instructions.

See Preparing Datasets for CSL.

See Getting Started with CSL.

Shield:

The majority of CSL is licensed under a MIT License.

However portions of the project are available under separate license terms: Swin-Transformer-Semantic-Segmentation is licensed under the MIT license, Deformable-DETR is licensed under the Apache-2.0 License.

If you find the code useful, please also consider the following BibTeX entry.

@misc{zhang2023csl,
      title={CSL: Class-Agnostic Structure-Constrained Learning for Segmentation Including the Unseen}, 
      author={Hao Zhang and Fang Li and Lu Qi and Ming-Hsuan Yang and Narendra Ahuja},
      year={2023},
      eprint={2312.05538},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Code is largely based on MaskFormer (https://github.com/facebookresearch/MaskFormer) and Mask2Former (https://github.com/facebookresearch/Mask2Former).