An Improved Traffic Sign Recognition Algorithm Based on Deep Learning

Aiming at the disadvantages of convolutional neural network in traffic sign recognition, such as poor real-time performance and high hardware requirements, an improved network based on lightweight convolutional neural network with real-time performance and high precision is proposed. On the one hand, depthwise separable convolution and activation function Mish are introduced to speed up the network training and recognition speed and reduce the requirements for hardware devices; on the other hand, through the improvement of network architecture and level, the size and number of convolution cores are reasonably changed to enhance the expression and transmission of image features. The experimental results on the traffic sign dataset of Belgium TSC show that the improved network significantly improves the network training speed, and the recognition accuracy is slightly higher than that of the original network, which verifies the effectiveness of the improved method. Compared with other models, this model can complete the task of traffic sign recognition more quickly and accurately, which verifies the feasibility of this method.