Research on Chinese Minority Clothing Based on Deep Convolution Neural Network

There are 55 minority nationalities in China. The traditional clothing of minority nationality is a part of Chinese traditional culture. How to use computer technology to identify these clothing has great significance for the protection and inheritance of Chinese traditional culture. Some scholars have studied the methods of minority nationality clothing recognition, but these methods require people to manually mark the semantic attributes of clothing on the training set pictures. In this paper, an end-to-end method of deep convolution neural network is adopted to implement the Chinese minority nationality clothing classifier. On the basis of CNN pre-training model mobilenet-v2, Fine-tune training can classify minority nationality clothing without manual marking on the training set pictures, and the accuracy rate is 94.5%.