[TGRS2026] [FreMaNet] Lightweight ORSI Salient Object Detection via Frequency and Mutual Assistance Attention IEEE Link|PDF|Homepage

python 3.8 + pytorch 1.13.1

We provide saliency maps of our FreMaNet, lightweight methods (code: frem), and normal-size methods (code: frem) on the ORSSD, EORSSD, and ORSI-4199 datasets.

We use data_aug.m for data augmentation.

Modify paths of datasets, then run train_FreMaNet.py.

Note: Our main model is under './model/GeleNet_models.py'. Our code is built on GeleNet. So in this code, GeleNet refers to our FreMaNet.

1. We provide the pre-trained models in './models/'.

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

3. Run test_FreMaNet.py.

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

    @ARTICLE{Li_2026_FreMaNet,
            author = {Gongyang Li and Shixiang Shi and Yong Wu and Weisi Lin and Zhen Bai},
            title = {Lightweight ORSI Salient Object Detection via Frequency and Mutual Assistance Attention},
            journal = {IEEE Transactions on Geoscience and Remote Sensing},
            volume = {},
            year = {2026},
            }

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.