An improved method MSS-YOLOv5 for object detection with balancing speed-accuracy

For deep learning-based object detection, we present a superior network named MSS-YOLOv5, which not only considers the reliability in complex scenes but also promotes its timeliness to better adapt practical scenarios. First of all, multi-scale information is integrated into different feature dimensions improve distinction and robustness features. The design detectors increases variety detection boxes accommodate wider range detected objects. Secondly, pooling method upgraded obtain more detailed information. At last, add Angle cost assign new weights loss functions accelerate convergence accuracy detection. In our network, explore four variants MSS-YOLOv5s, MSS-YOLOv5m, MSS-YOLOv5x, MSS-YOLOv5l. Experimental results MSS-Yolov5s show that technique improves mAP on PASCAL VOC2007 2012 datasets by 2.4% 2.9%, respectively. Meanwhile, it maintains fast inference speed. same time, other three models have degrees performance improvement terms balancing speed precision challenging regions.