A face authentication method and device

Abstract

Embodiments of the present invention provide a face authentication method and device, the method comprising: dividing the target face image in the target image into a plurality of segmented face images; Input in the preset neural network model, extract corresponding multiple original facial features; Select at least part of the original facial features to generate target facial features from the multiple original facial features; The target face image is authenticated. The embodiment of the present invention adopts a single network, multi-block method, and a training strategy in which the network and the classifier are separated, so the number of training samples is small, which reduces the possibility of overfitting during the training process. No end-to-end connection is used to construct higher-dimensional features, and the target face features with the same dimension as the original face features, which are more conducive to recognition and more robust to changes are constructed.

Description

Technical field [0001] The present invention relates to the technical field of biological data, and more particularly to a face certification method and a face certification device. Background technique [0002] Since face certification is natural and not subject to the characteristics of being detected, face certification is widely used in financial and information security. [0003] At present, the face certification protocol generally has two traditional machine learning methods and two kinds of deep learning methods. It can be divided into a method based on deep learning and metrics, based on deep learning and metrics. Separation training method. [0004] Based on traditional machine learning methods, most is human-oriented design features, and extract characteristics from face images by statistical learning methods, and further face certification. [0005] However, since people's subjective ideas tend to have certain sideways, the extraction is often only tends to have robus...

AI Technical Summary

Problems solved by technology

[0009] In view of the above problems, in order to solve the above problems of one-sided features, high multi-network complexity, and high dimensionality of feature fusion, an embodiment of the present invention proposes a face authentication method and a corresponding face authentication device

Images