Home

Awesome

ACCoNet

This project provides the code and results for 'Adjacent Context Coordination Network for Salient Object Detection in Optical Remote Sensing Images', IEEE TCYB, 2023. IEEE link and arxiv link Homepage

Network Architecture

<div align=center> <img src="https://github.com/MathLee/ACCoNet/blob/main/image/ACCoNet.png"> </div>

Requirements

python 2.7 + pytorch 0.4.0 or

python 3.7 + pytorch 1.9.0

Saliency maps

We provide saliency maps of our ACCoNet (VGG_backbone (code: gr06) and ResNet_backbone (code: 1hpn)) on ORSSD, EORSSD, and additional ORSI-4199 datasets.

Image

Training

We provide the code for ACCoNet_VGG and ACCoNet_ResNet, please modify '--is_ResNet' and the paths of datasets in train_ACCoNet.py.

For ACCoNet_VGG, please modify paths of VGG backbone (code: ego5) in /model/vgg.py.

data_aug.m is used for data augmentation.

Pre-trained model and testing

  1. Download the following pre-trained models and put them in /models.

  2. Modify paths of pre-trained models and datasets.

  3. Run test_ACCoNet.py.

ORSSD: ACCoNet_VGG (code: 1bsg); ACCoNet_ResNet (code: mv91).

EORSSD: ACCoNet_VGG (code: i016); ACCoNet_ResNet (code: ak5m).

ORSI-4199: ACCoNet_VGG (code: qv05); ACCoNet_ResNet (code: art7).

Evaluation Tool

You can use the evaluation tool (MATLAB version) to evaluate the above saliency maps.

ORSI-SOD_Summary

Citation

    @ARTICLE{Li_2023_ACCoNet,
            author = {Gongyang Li and Zhi Liu and Dan Zeng and Weisi Lin and Haibin Ling},
            title = {Adjacent Context Coordination Network for Salient Object Detection in Optical Remote Sensing Images},
            journal = {IEEE Transactions on Cybernetics},
            volume = {53},
            number = {1},
            pages = {526-538},
            year = {2023},
            month = {Jan.},
            }
            
            

If you encounter any problems with the code, want to report bugs, etc.

Please contact me at lllmiemie@163.com or ligongyang@shu.edu.cn.