This is an implementation of "PolSAR Image Classification using Shallow to Deep Feature Fusion Network with Complex Valued Attention" Accepted for Publication on Scientific Reports. The paper can be accessed through:
https://www.nature.com/articles/s41598-025-10475-3
Three benchmark datasets were used in this paper, namely Flevoland, San Francisco and Oberpfaffenhofen, dataset can be downloaded from: https://mega.nz/folder/WhgT1L4S#PnMttCUpjtwkD8qTEdwZsw
Python 3.9.18, Tensorflow (and Keras) 2.10.0, cvnn 2.0, Tensorflow Probability 0.18.0
To quantitatively measure the proposed CV-ASDF2Net model, three evaluation metrics are employed to verify the effectiveness of the algorithm, Overall Accuracy (OA), Average Accuracy (AA) and Cohen's Kappa (k). Also, Each class accuracy has been reported
Model was qualitatively evaluated by visually comparing the resulting class maps.
@article{alkhatib2025polsar, title={PolSAR image classification using shallow to deep feature fusion network with complex valued attention}, author={Alkhatib, Mohammed Q and Zitouni, M Sami and Al-Saad, Mina and Aburaed, Nour and Al-Ahmad, Hussain}, journal={Scientific Reports}, volume={15}, number={1}, pages={24315}, year={2025}, publisher={Nature Publishing Group UK London} }