Awesome

Feature Shrinkage Pyramid for Camouflaged Object Detection with Transformers (CVPR2023)

<p align="left"> <a href="https://arxiv.org/abs/2303.14816"><img src="https://img.shields.io/badge/Paper-arXiv-green"></a> <a href="https://tzxiang.github.io/project/COD-FSPNet/index.html"><img src="https://img.shields.io/badge/Page-Project-blue"></a> </p>

Usage

The training and testing experiments are conducted using PyTorch with 8 Tesla V100 GPUs of 36 GB Memory.

1. Prerequisites

Note that FSPNet is only tested on Ubuntu OS with the following environments.

Creating a virtual environment in terminal: conda create -n FSPNet python=3.8.
Installing necessary packages: pip install -r requirements.txt

2. Downloading Training and Testing Datasets

Download the training set (COD10K-train) used for training
Download the testing sets (COD10K-test + CAMO-test + CHAMELEON + NC4K ) used for testing

3. Training Configuration

The pretrained model is stored in Google Drive and Baidu Drive (xuwb). After downloading, please change the file path in the corresponding code.
Run train.sh or slurm_train.sh as needed to train.

4. Testing Configuration

Our well-trained model is stored in Google Drive and Baidu Drive (otz5). After downloading, please change the file path in the corresponding code.

5. Evaluation

Matlab code: One-key evaluation is written in MATLAB code, please follow this the instructions in main.m and just run it to generate the evaluation results.
Python code: After configuring the test dataset path, run slurm_eval.py in the run_slurm folder for evaluation.

6. Results download

The prediction results of our FSPNet are stored on Google Drive and Baidu Drive (ryzg) please check.

Citation

@inproceedings{Huang2023Feature,
title={Feature Shrinkage Pyramid for Camouflaged Object Detection with Transformers},
author={Huang, Zhou and Dai, Hang and Xiang, Tian-Zhu and Wang, Shuo and Chen, Huai-Xin and Qin, Jie and Xiong, Huan},
booktitle={IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2023}
}

Thanks to Deng-Ping Fan, Ge-Peng Ji, et al. for a series of efforts in the field of COD.