A Yolov3-Based Multi-target Detection System for Complex Scenes

Abstract

With the rise of deep learning technology in recent years, many deep learning techniques have been applied to several aspects, and a deep learning-based target detection system is one of them. In fact, traditional target detection for complex scenes usually faces many problems, including spatial occlusion, small target and multi-target detection, and real-time detection efficiency. In response to this phenomenon, this paper adopts the YOLOv3 algorithm and uses the Pascal VOC2007 dataset for model training to build a multi-target detection system. The experimental results show that YOLOv3 can still detect objects in complex scenes with classical dataset training, and mitigate the effect of spatial occlusion on target detection compared with traditional target detection algorithms, which has a certain application value for complex scenes.