Swin Transformers for Remote Sensing SAR Image Classification

The Swin Transformer, introduced in 2021, has gained widespread recognition for its ability to effectively capture long-range dependencies and contextual information in image data. This project aims to leverage the success of the Swin Transformer architecture in various domains by applying it to Synthetic Aperture Radar (SAR) image classification, with a particular focus on the renowned MSTAR (Moving and Stationary Target Acquisition and Recognition) dataset. The initiative begins with a comprehensive pre-processing phase tailored to optimize SAR data compatibility with the Swin Transformer. This involves meticulous steps such as data formatting, radiometric calibration, noise reduction, and feature extraction, all customized to the unique characteristics of SAR imagery present in the MSTAR dataset. The work serves as a testament to the fusion of state-of-the-art transformer technology with the intricacies of SAR imagery. The ultimate objective is to enhance the accuracy of SAR image classification, thereby improving its applications in crucial areas like remote sensing and defense. This endeavor highlights the potential synergy between cutting-edge transformer architectures and the complexities inherent in SAR data processing, paving the way for advancements in remote sensing technologies.