Glasses removal model training method, face recognition method, device and equipment

Abstract

The invention discloses a training method of a glasses removal model, a face recognition method and a face recognition device. The training method comprises the steps that a sample image is acquired,the sample image comprises an input image and a target image, the difference between the input image and the target image is whether a human face in the image wears glasses or not, the input image isused for indicating a human face image wearing glasses, and the target image is used for indicating a human face image not wearing glasses; a symmetric convolutional neural network model is constructed, the model adopts a symmetric convolutional neural network structure and comprises an input layer, N convolutional layers, N deconvolutional layers and an output layer which are sequentially connected, and the ith convolutional layer and the (N+1-i)th deconvolutional layers are in a symmetric relationship, the convolution layers are used for carrying out feature coding on the image, and the deconvolution layers are used for carrying out feature decoding on the image; and the model is trained according to the input image and the target image to obtain a glasses removal model. The method can enable the trained model to have a better effect.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a training method for a glasses removal model, a face recognition method, a device, an electronic device, and a computer-readable storage medium. Background technique [0002] The eye area is the main feature area of ​​the human face, and glasses are common facial occlusions, especially dark thick-rimmed glasses have obvious features, which greatly interfere with the feature extraction of eye images. People with similar glasses in face recognition are easily misidentified, or a passport photo without glasses and a life photo with glasses are difficult to identify as the same person. Therefore, removing glasses is an important step in face recognition. [0003] In the related art, the following centralized methods are commonly used to remove the glasses: 1) detect the outline of the glasses based on edge detection, and then fuse the surrounding pixels to estimate...

AI Technical Summary

Problems solved by technology

[0004] However, the current problem is that the method based on edge detection is easily affected by the type of glasses and lighting. For example, it is difficult to detect edge information for frameless or half-rim glasses, and it is difficult to perform pixel fusion when there is reflection; based on Although the reconstruction method of PCA is simple, it is based on the projection and reconstruction of the entire face. Not only will there be residue in the glasses area, but also noise will be introduced in other areas of the face, and the effect is poor.

Images