YOLOX-based ship target detection for Shore-based monitoring

With the development of artificial intelligence technology, people have higher requirements for intelligent monitoring system in terms of accuracy and detection speed of ship target detection algorithm. YOLOX, as the latest proposed high-performance real-time detection model, adds advanced target detection techniques such as decoupling head and tag assignment strategy SimOTA to the YOLO series, and achieves very good results on the COCO dataset. Therefore, we use it on a ship intelligent monitoring system to study its effectiveness on ship target detection tasks. We evaluate the performance of YOLOX using the well-known ship image dataset Seaships and compare the results with those of other studies. The results show that YOLOX achieves the best performance on the dataset. Among them, YOLOX_Tiny(The easiest YOLOX) has the fastest detection speed with an FPS of 129.3; YOLOX_X(The most complex YOLOX) achieves 97.54% of mAP. this is the best performance on this dataset so far.