A face model training method based on Center Loss improvement

Abstract

The invention relates to a face model training method based on Center Loss improvement, and the method comprises the steps: (1),cutting and screening face images in an original total data set, and dividing the face images into a training set, a verification set, and a test set; (2) preprocessing the face images in the training set; (3) building a network structure, and optimizing a target loss function; (4) inputting the data in the training set into a network structure for training; (5) storing the face model; and (6) testing the face model by using the test set. On the basis that the face model has a certain classification capability, the intra-class aggregation capability of the model is enhanced, the targets of effectively increasing the inter-class distance and reducing the intra-class distance are achieved, and the accuracy and robustness of face recognition are improved.

Description

technical field [0001] The invention relates to an improved face model training method based on Center Loss, which belongs to the technical field of deep learning. Background technique [0002] With the continuous development of society and the urgent need for fast and effective identity verification in all aspects, people pay more and more attention to the research of face recognition, and the accuracy requirements of face models are also getting higher and higher. Face recognition technology is based on the facial features of a person, giving the position and size of the input face image and the location information of each major facial organ, and further extracting the features of each face and comparing them with known faces to identify people. identity of. Compared with other biotechnology, face recognition has the advantages of non-contact, non-mandatory, and concurrency, so it plays an irreplaceable role in identity verification, security monitoring, etc. [0003] A...

AI Technical Summary

Problems solved by technology

In machine learning, especially deep learning, softmax is a very important function, especially widely used in multi-classification scenarios, and for common image classification problems, SoftMax Loss is often used to find losses, but, based on the traditional sense The supervision of the SoftMax Loss cost function of the network structure, the category boundary is clear and the intra-class difference is too large

Some scholars have proposed to combine SoftMax Loss and Center Loss as the loss function of neural network training, improve the core idea of ​​the algorithm, enhance inter-class differences and reduce intra-class differences. However, the design of the above loss function is only for training. The proportion of Center Loss is fixed in the process, and the degree of intra-class aggregation cannot be adjusted according to the training parameters. It is easy to fall into a local optimal solution and lacks flexibility.

Images