关键词: 神经网络, 迁移学习, 交通标志, 图像识别

Abstract: In view of time-consuming training and low accuracy of the traditional traffic sign recognition algorithms, this paper proposes a traffic sign recognition model based on ResNet and transfer learning. Firstly, the ResNet network weight trained on ImageNet image data set is introduced into the model, the parameters of the convolutional layer are frozen, and the network is used as the feature extractor of the model. Secondly, a fully connected layer is designed for the model, and the parameters of the fully connected layer are fine-tuned by using the data sets of different sizes and the data sets before and after data augmentation. Then, different sizes of learning rates are set, and the model is trained under two conditions of fixed learning rate and decaying learning rate. Finally, the model is tested on the test set and the classification results are output.The test results show that the recognition accuracy of traffic signs by this method is 97.60%, and the F1 score of the three classes of traffic signs reach 96.86%, 99.37% and 96.53% respectively, including mandatory signs, warning signs and prohibition signs, which shows that the model has a high recognition accuracy of traffic signs.