Chinese Traditional Visual Cultural Symbols recognition based on Convolutional neural network

Chinese Traditional Visual Cultural Symbols(CT-VCSs) is the important component of Chinese ancient civilization, and it is the crystallization of Chinese culture with a history of several thousand years. So it has great significance to research CT-VCSs. In this paper, we mainly research the recognition and classification of CT-VCSs based on Convolutional neural network(CNN). We mainly use Caffenet and Alexnet in the Caffe framework, and fine-tune the existed Caffe models. Meanwhile, we also use GPU to speed up the process of training. Experimental results indicate that using CNN poses remarkable enhancement on the recognition task of CT-VCSs, and the recognition result of using Alexnet is the best.