TSCNet

This project provides the code and results for 'Texture-Semantic Collaboration Network for ORSI Salient Object Detection', IEEE TCAS-II, 2024. IEEE link and arxiv link, Homepage

Network Architecture

Requirements

python 3.8 + pytorch 1.9.0

Saliency maps

We provide saliency maps saved using two different functions (imageio.imsave and cv2.imwrite) on ORSSD, EORSSD, and ORSI-4199 datasets.

Using "imageio.imsave(res_save_path+name, res)" to save saliency maps in "./models/saliencymaps_imageio.zip"(code: 6sr5), termed TSCNet_imageio in Table I (reported in our paper).

Using "cv2.imwrite(save_path+name, res*256)" to save saliency maps in "./models/saliencymaps_cv2.zip", termed TSCNet_cv2 in Table I.

Training

We use data_aug.m for data augmentation.

Download VGG weight (code: ipbb), and put it in './model/'.

Download ViT weight (code: nd45), and put it in './network/'.

Run train_TSCNet.py.

Pre-trained model and testing

Download the following pre-trained model, and modify paths of pre-trained model and datasets, then run test_TSCNet.py.

ORSSD (code: t6it)

EORSSD (code: f9dv)

ORSI-4199 (code: jcm8)

Evaluation Tool

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

ORSI-SOD_Summary

Citation

    @ARTICLE{Li_2024_TSCNet,
            author = {Gongyang Li and Zhen Bai and Zhi Liu},
            title = {Texture-Semantic Collaboration Network for ORSI Salient Object Detection},
            journal = {IEEE Transactions on Circuits and Systems II: Express Briefs},
            volume= {71},
            number={4},
            pages={2464-2468},
            year={2024},
            month={Apr.},
            }

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.