Mask Face Recognition Method Based on Double-Branch Weight Fusion Homology Self-Supervision

Abstract

This application relates to a self-supervised mask face recognition method based on dual-branch weight fusion homology. The face recognition model obtains a well-trained underlying feature sharing network model; the homology supervision loss function and the face recognition Arcface loss loss function are fused by dual branch weights, respectively for the left branch sub-model and the right branch sub-model in the face recognition dual branch model. The branch sub-model is trained, and the trained left branch sub-model and the right branch sub-model are weight-fused to obtain a high-level semantic fusion network model; the trained low-level feature sharing network model is spliced ​​with the high-level semantic fusion network model to obtain The final face recognition prediction model. Through this application, the industry problem of how to improve the pass rate of mask faces is solved, the pass rate of mask face recognition is greatly improved, the recognition accuracy is improved, and the cost is reduced.

Description

technical field [0001] This application relates to the technical field of face recognition, in particular to a method for mask face recognition based on double-branch weight fusion homology self-supervision. Background technique [0002] Nowadays, wearing a mask as one of the important means of epidemic prevention has become the norm in people's life and work. Under the premise of ensuring the same misrecognition rate, how to improve the pass rate of masks and faces is still a difficult problem in the industry. For example, when masks cover most of the face, the general-purpose face recognition system can no longer accurately and quickly meet the traffic needs of airport real-name ticket check-in, company attendance or community access control systems, and provide city information related to face recognition technology. The system brings new challenges. Therefore, be badly in need of a kind of scheme that solves the above problems. [0003] In fact, the face image recognit...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a mask face recognition method based on dual-branch weight fusion homology self-supervision, so as to at least solve the problem of low recognition pass rate in the related art when mask faces are recognized

Images