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.

COD datasets: download the COD datasets set from here(CAMO, CHAMELEON, COD10K, NC4K), and put into 'data/'

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.

the generated boundary image dataset:链接：https://pan.baidu.com/s/14b9LYOYaTQ0CQ7qukbWLtg 提取码：vmlp

python pre_npz_UEDG.py

Weights

pre-weigth: download the weight of sam from here, put into 'work_dir_cod/SAM/'
COMPrompter: download the weight of well-trained COMPrompter, put into 'work_dir_cod/COMPrompter'
- 链接：https://pan.baidu.com/s/1Hr7q7x3bctnEtdS7XeciUw 提取码：dv67

The predicted image

COMPrompter using the perfect boundary:
- 链接：https://pan.baidu.com/s/1MQDx1bVKoYCPFSiFV9XVew 提取码：u9cj
COMPrompter using the generated boundary:
- 链接：https://pan.baidu.com/s/1ezx5cl2T4hCe2O55I8Dygw 提取码：l051

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>

}