Face recognition method and device based on deep learning

Abstract

The invention discloses a face recognition method and device based on deep learning, and the method comprises the steps: obtaining a face image training sample and a to-be-detected face image; extracting face training features from the face picture training sample, and extracting to-be-detected face features from the to-be-detected face picture; constructing a convolutional neural network model, and training the face training features of the face picture training sample by using the convolutional neural network model to obtain a face recognition model; and comparing the to-be-detected face features of the to-be-detected face picture according to the trained face recognition model so as to recognize the to-be-detected face picture. According to the method of the invention, the method can enable the inter-class spacing to be more uniform, can train more types of training data, can achieve the large-scale face data training, can improve the face recognition efficiency, and improves the face recognition performance.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a face recognition method, device, electronic equipment and readable storage medium based on deep learning. Background technique [0002] With the development of face recognition technology, various face recognition related products have been widely used in people's lives. At present, the main recognition function of face recognition technology is realized based on Convolutional Neural Network (CNN). Using a large number of face picture data sets, the convolutional neural network is trained to have the ability of face recognition after the convolutional neural network training converges. Considering that many current products need to target millions of person identities, the difficulty of training network models increases. To this end, many current training methods use the classification activation function softmax to define the identity, so that the process of...

AI Technical Summary

Problems solved by technology

But it exists: 1. When training the loss function of RegularFace, if there are more training samples in a certain category, it may cause greater interference in the distance between classes, making the distance between categories not uniform enough; 2. In the early stage of training , because the model does not form a good classification function, that is, the center points of each category represented by the W parameter of the convolutional layer are not sufficiently separated, which will lead to an extension of the training time; 3. When the number of training sample categories is large, The calculated amount of cosine distance between classes will be very large, making it difficult or impossible for most current computers to run

Images