TASE-Adaptive Task Alignment and Semantic Enhancement Network for High-Altitude UAV Target Detection

In recent years, the application of drones in daily life has become increasingly widespread, leading to a surge in demand for aerial target identification tasks using drones. However, when drones observe targets from high altitudes, they not only encounter environmental interferences such as buildings and trees, but also face a broad observation range, resulting in small sizes and diverse categories of targets of interest. Traditional target detection methods exhibit certain limitations in this scenario. To address these issues, this paper proposes an adaptive task alignment and semantic enhancement network for aerial target detection using drones, based on YOLOv8s. We improve the neck and detection head parts of the original algorithm and attempted to use the latest network structure modules in the backbone network. Finally, we conduct experiments on the VisDrone2019-DET dataset, and the experimental results demonstrate the effectiveness of the proposed algorithm.