Abstract
This study proposes a novel transfer learning framework for effective shipclassification using high-resolution optical remote sensing satellite imagery.The framework is based on the deep convolutional neural network model ResNet50and incorporates the Convolutional Block Attention Module (CBAM) to enhanceperformance. CBAM enables the model to attend to salient features in theimages, allowing it to better discriminate between subtle differences betweenships and backgrounds. Furthermore, this study adopts a transfer learningapproach tailored for accurately classifying diverse types of ships byfine-tuning a pre-trained model for the specific task. Experimental resultsdemonstrate the efficacy of the proposed framework in ship classification usingoptical remote sensing imagery, achieving a high classification accuracy of 94%across 5 classes, outperforming existing methods. This research holds potentialapplications in maritime surveillance and management, illegal fishingdetection, and maritime traffic monitoring.