An Improved YOLOv3 Model Based on ResNet50

In computer vision, object detection is one of the most challenging problems. In the one-stage object detection, YOLOv3 is one of the most popular object detection models. Because of its excellent detection speed and the same accuracy, it has a wide range of practical applications, and many kinds of derivative versions have been derived. In this paper, we used four tricks to upgrade the original yolov3 network which are resnest50, image mixup, label smoothing and CIoU loss. By adding these tricks, the average accuracy and speed of the improved model on coco2017 datasets are better than those of traditional YOLOv3.