Implementation and Evaluation of CNN Based Traffic Sign Detection with Different Resolutions
Ying-Chi Chiu, Huei‐Yung Lin, Wen-Lung Tai · 2019
In this paper, we develop a traffic sign detection and recognition technique based on the YOLOv3 framework. The public traffic dataset TT100K and our own Taiwan road scene image dataset are used for network training. In the performance evaluation, we have extensively compared different input image resolution and ROI for network training and testing. In the experiments, the proposed method is able to achieve the mAP of 92.99% with 21 most common traffic signs on the road.