在研究中，结合申请者已有工作成果，提出基于VGG-Face2的多尺度集

In the study, combined with the results of the work of the applicant has been proposed based on multi-scale VGG-Face2 integrated feature construction method, and using a combination of distinction between learning and generate learning in depth study, the distinguishing feature type and generative characteristics integration, and finally through probabilistic modeling approach to achieve recognition matching face.

In the study, combined with the applicant's existing work results, the construction method of multi-scale integration features based on VGG-Face2 is proposed, and the differentiated and generative learning is used in deep learning, the differentiated features will be integrated with the generative features, and finally the recognition matching of faces is achieved through the probability model method.

In the research, combined with the existing work of the applicant, a multi-scale integrated feature construction method based on vgg-face2 is proposed. In-depth learning, differentiated learning and generative learning are integrated to integrate the differentiated feature and the generative feature. Finally, the probability model method is used to achieve face recognition and matching.