Facial Ethnicity Recognition Based on Transfer Learning from Deep Convolutional Networks

With the development of deep learning, computer face recognition has made significant progress. However, face ethnic characteristics information is rarely used in face recognition technology. The research of facial ethnicity recognition had not only been directly applied in daily life, but also avoided racial effects and improved model generalization performance. In the paper, we proposed a Chinese facial ethnicity recognition (CFER) model based on transfer learning from deep convolution networks. First, we collected 5 Chinese ethnic groups to build a face dataset containing ethnicity information; then we have applied CFER to recognize Chinese ethnicity characteristics and 10-fold cross validation method to estimate mainly the accuracy rate of the model. The average recognition rate of the model is 80.5%, meanwhile, the model also has good generalization performance. It's proved that deep learning method is feasible for facial ethnicity recognition.