基于深度卷积神经网络的车标分类

Vehicle classification based on deep convolutional neural network

基于深度学习框架Caffe和具有强大计算能力的GPU,运用深度神经网络AlexNet和GoogleNet,对具有不同背景的汽车图像进行网络训练,以达到车辆自动识别的目的。分别对4类车标进行网络训练与测试,实验结果表明,在图像识别分类方面,与传统识别方法相比,深度卷积神经网络更具优势。

Based on Caffe of deep learning frame and GPU with power computing capacity, using deep neural network AlexNet and GoogleNet, network training is carried out on vehicle images with differ-ent background for automatic recognition of vehicle. In this paper, network training and tests on are car-ried out on four types of auto logo. The experiments show that in terms of image recognition and classifi-cation, deep convolutional neural network has more advantages compared to conventional method.