Traffic Scene Detection Based on YOLOv3

With the development of urbanization and the ongoing big data scene, intelligent object detection in traffic scenes has become a hot issue at the moment. In this paper, the largest computer vision algorithm evaluation data set in the world, the KITTI data set, is trained on the model of YOLOv3. K-means method is used to adjust anchor parameters to achieve more accurate results. The experimental results show that the traffic scene detection model trained in this paper has good detection effect.