The official implementation of our AAAI 2025 publication SPT.

conda create -n spt python=3.8
conda activate spt
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
pip install matplotlib
pip install opencv-python
pip install timm
pip install scikit-image
pip install imgaug
pip install pandas

Please download the dataset using this link and extract the archive directly into the repository root.

• Pre-trained SAM checkpoints. Download here

  pretrained_checkpoint/
  ├── vit_b.pth
  ├── vit_h.pth
  └── vit_l.pth
  

• SPT checkpoints. Download here

  spt_ckpt/
  ├── spt_vit_b.pth
  ├── spt_vit_h.pth
  └── spt_vit_l.pth
  

To launch inference in one step, simply run:

sh main.sh

If you find our work helpful for your research, please consider citing:

@inproceedings{yang2025promptable,
  title={Promptable anomaly segmentation with sam through self-perception tuning},
  author={Yang, Hui-Yue and Chen, Hui and Wang, Ao and Chen, Kai and Lin, Zijia and Tang, Yongliang and Gao, Pengcheng and Quan, Yuming and Han, Jungong and Ding, Guiguang},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={39},
  number={12},
  pages={13017--13025},
  year={2025}
}

This codebase is built upon SAM, LoRA, HQ-SAM, Grounded SAM and MobileSAM

Thanks for their public code and released models.