Future fishery and deep learning

Ship classification algorithms based on the analysis of remote sensing data, e.g. Synthetic Aperture Radar (SAR) images, are examples of data processing methodologies that fulfill a crucial step in monitoring the marine environment. We compared the potential of different Deep Learning architectures with Vision Transformers (ViTs), a recent advancement in deep learning introducing an attention-based mechanism, for ship classification on imbalanced SAR data using F1-Score as metric.