LAUSR

Existing anchor-based rotated object detection methods have achieved some amazing results, but these methods require some manual preset anchors, which not only introduce additional hyperparameters but also introduce extra computational burdens. Due to the above drawbacks, anchor-free methods have been rapidly developed in recent years. However, the existing high-performance anchor-free rotated object detection methods are relatively complex and the inference speed is also slow. And you only look once detector (YOLO) series models not only maintain high-efficiency inference but also keep competitive performance detection performance in the general object detection tasks. Hence, we propose an anchor-free rotated detector based on the exceeding you only look once detector (YOLOX) method for object detection in aerial images. Our methods consist of two improvements: a refined rotated module (RRM) and a new assigner method which is called the Gaussian distribution sampling optimal transport assignment (GSOTA) method. The RRM can align features and get more useful priors for final detector heads. The GSOTA uses Gaussian distribution to model the oriented bounding box (OBB) first, and a Gaussian center sampling (GCS) method with maximum classification center mean (MCCM) is proposed to simplify the label Assignment optimal transport (OT) problem, finally using an improved dynamic top- $k$ strategy to get an approximate solution. Extensive experiments demonstrate that our models can achieve competitive performance in several challenging aerial object detection datasets while keeping the best efficiency. Our R2YOLOX-X model achieves 79.33%, 97.4%, 97.7%, and 92.5% mean of Average Precision (mAP) on the dataset for object detection in aerial image (DOTA), HRSC2016, University of Chinese Academy of Science (UCAS)-dataset of object detection in aerial image (AOD), and FGSD2021, respectively, while R2YOLOX-S can reach the fastest 58.2 FPS when inferencing on aerial datasets and R2YOLOX-L gets the best speed-accuracy trade-off.