This is an implementation of "PolSAR Image Classification Using Attention Based Shallow to Deep Convolutional Neural Network" that was published in IEEE-IGARSS 2023

To validate this statement, the performance of the proposed network is evaluated on the Felovland PolSAR image. The scene was acquired by the NASA/JPL AirSAR system over the agricultural area in Netherlands with a size of 750x1024. The number of labeled samples is 207,832.

python 3.9, Tensorflow 2.10.0, Spyder IDE, CVNN package

To quantitatively measure the proposed model, three evaluation metrics are employed to verify the effectiveness of the algorithm, including Overall Accuracy (OA), Average Accuracy (AA) and Cohen's Kappa (k).

Model was qualitatively evaluated by visually comparing the resulting class maps.

@INPROCEEDINGS{10282338, author={Alkhatib, Mohammed Q. and Al-Saad, Mina and Aburaed, Nour and Zitouni, M. Sami and Al-Ahmad, Hussain}, booktitle={IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium}, title={PolSAR Image Classification Using Attention Based Shallow to Deep Convolutional Neural Network}, year={2023}, volume={}, number={}, pages={8034-8037}, doi={10.1109/IGARSS52108.2023.10282338}}

Feel Free to contact me on: [email protected]