Abstract: Aiming at the problem of accuracy of vehicle logo recognition, we proposed a vehicle logo recognition algorithm based on improved residual network of ResNet-18 model. Firstly, the residual network was used and improved, and the improved linear correction unit Leaky ReLU activation function was used to replace the original activation function. Secondly, we adjusted the traditional residual network structure, put the batch standardization and activation function before the convolution layer, and reduced the network parameters to speed up the network training. The experimental results show that the recognition accuracy of the improved residual network model is 99.8%.

Key words: deep learning, residual network, image recognition, vehicle logo recognition