Face recognition model training method, device, system and computer readable medium

Abstract

The invention provides a face recognition model training method, a device, a system and a computer readable medium. The training method of the face recognition model comprises the following steps: inputting N groups of pictures into a neural network in each batch to obtain a feature vector corresponding to each picture, each group of pictures in the N groups of pictures belonging to the same category, and N being a natural number greater than or equal to 1; calculating the intra-class distance of each group of pictures based on the feature vector, and calculating a first loss function according to the intra-class distance for monitoring the distribution difference of the intra-class distance; calculating a second loss function, and weighting the second loss function and the first loss function to obtain a total loss function; and optimizing the total loss function to converge the total loss function. According to the method, an intra-class distance distribution difference loss functionis introduced in the training process, intra-class distance distribution is normalized, and the problem that thresholds are difficult to unify due to different data set distribution differences can be solved.

Description

technical field

[0001] The present invention relates to the technical field of face recognition, and more particularly to a training method, device, system and computer-readable medium of a face recognition model. Background technique

[0002] The current face recognition tasks are mainly divided into three categories, namely face verification (verify whether it is the same person), face recognition (find the person who is closest to the query face picture and many target face pictures) and clustering (target face images, grouping them into those that look the most like each other). The usual method is to transform the face picture into a point in the feature space by training a deep network model, and make the face corresponding to the closest point in the feature space most resemble the same person, and the farther point corresponds to a different person. . Then, the face verification task is equivalent to calculating whether the distance between points in the feature sp...

Images