Traffic Sign Detection Based on YOLO v3
Xilin Zhang · 2023
Machine vision has penetrated into every field of our daily life and is being used more and more widely, especially in the field of automatic driving in recent years, computer vision in ImageNet Dataset will show very strong accuracy and robustness. The research objective of this paper is to analyze the network structure of YOLO V3 and summarize the loss function, use the GTSDB dataset to train and test the YOLO V3 algorithm, get the recall rate, accuracy and average accuracy of YOLO V3 in identifying traffic signs, and compare the results with R-CNN series algorithms. After testing, we find that YOLO algorithm is faster and more accurate than RCNN algorithm in traffic sign recognition, which is exactly what automatic driving needs. This provides a new way for the traffic sign recognition algorithm in the future, and also has potential application value.