Abstract: To address the shortcomings of traditional deep learning applications for multi-label satellite image classification, this work shifts focus toward lightweight Vision Transformers (ViTs) ...