Abstract
In this work, we propose a two-branch neural network architecture for multi-label classification in UAV imagery. Compared to single-label classification, the multi-label classification problem aims to assign multiple class labels to the image, which is more challenging. To deal with this issue, the proposed network optimizes in an end-to-end manner three loss functions related to image-label similarity, label discrimination, in addition to the number of labels present in the image. The experiments carried out on two UAV datasets with a spatial resolution of 2-cm confirm the promising capability of the proposed method.