Home

Awesome

<div align="center"> <h2>COMPrompter: Rethink SAM in Camouflaged Object Detection with Multi-Prompt Network</h2> Xiaoqin Zhang, Zhenni Yu, Li Zhao, Deng-Ping Fan, Guobao Xiao* 2024 </div>

 Comparison of our COMPrompter and other methods in COD

Usage

Installation

git clone https://github.com/guobaoxiao/COMPrompter
cd COMPrompter

From datasets to npz

For the train npz and test npz of the perfect boundary with gradient (corresponding Ours in Tab. 1). You can load down the COD datasets and run to make npz one by one.

python pre_npz.py

For the test npz of the generated boundary with gradient (corresponding Ours* in Tab. 1). You should load down the dataset of the generated boundary and run to make npz one by one.

python pre_npz_UEDG.py

Weights

The predicted image

Train

python Train.py

Test

python Inference.py

Translate npz to img

python transformer_nzp_2_gt.py

eval

python MSCAF_COD_evaluation/evaluation.py
@article{zhang2024COMPrompter,
  title={COMPrompter: Rethink SAM in Camouflaged Object Detection with Multi-Prompt Network},
  author={Zhang, Xiaoqin and Yu, Zhenni and Zhao,  Li and Fan, Deng-Ping and Xiao, Guobao},
  journal={SCIENCE CHINA Information Sciences (SCIS)},
  volume={1},
  pages={1--14},
  year={2024}
</h2>

}