Key feature area matching face recognition method based on a stacked hourglass network

Abstract

The invention relates to the technical field of computer vision recognition, and provides a key feature area matching face recognition method based on a stacked hourglass network, which comprises thefollowing steps: collecting a training set, and preprocessing the training set; Preprocessing the input face image; Inputting the picture into a stacked hourglass network for feature extraction, and outputting a face key point heat map and key point position information; Cutting a key area of the original picture, and selecting a triple from the training set; Performing feature extraction on the key area to obtain a feature map F; Inputting the feature map F into an embedded layer to obtain a label E; Calculating a ternary loss function according to the L2 norm of the feature map, and repeating the above steps until the ternary loss function converges; And inputting the to-be-identified face image into the trained stacked hourglass network and face identification module, and outputting anidentified tag E. According to the method, the stacked hourglass network is introduced for face recognition, the influence of non-key areas is eliminated, the face recognition effect is effectively improved, and high robustness is achieved.

Description

technical field [0001] The present invention relates to the technical field of computer vision recognition, and more specifically, to a face recognition method based on stacked hourglass network-based key feature region matching. Background technique [0002] In recent years, with the introduction of deep convolutional neural networks, people have applied it to face recognition tasks and achieved good results. A large part of this is due to the fact that deep convolutional neural networks can extract robust features. [0003] In the field of face key point detection, some relevant researchers proposed to use the stacked hourglass network for face key point location. This method first corrects the input face image, and then extracts multi-scale and discriminative features through the stacked hourglass network. , and finally perform regression positioning on the key points. In the prior art, a large number of algorithms are improved based on the above-mentioned stacked hourgl...

AI Technical Summary

Problems solved by technology

In the prior art, a large number of algorithms are improved based on the above-mentioned stacked hourglass network, but a challenging problem that still exists is that under different lighting conditions, different postures, and different expressions, the existing algorithms in the prior art are not as good as human beings. Face image input is not very robust

Images