Face recognition method, system and device based on centralized coordination learning

Abstract

The invention discloses a face recognition method, system and device based on centralized coordination learning, and the method comprises the following steps: obtaining a to-be-recognized face image,carrying out the face detection of the face image, and obtaining a first face image; After alignment processing is carried out on the first face image, a second face image with a preset size is obtained; inputting the second face image into a preset face recognition model based on centralized coordination learning for feature extraction, and obtaining a face feature vector of the second face image; and calculating cosine similarity by combining the face feature vector and a preset face database, and obtaining a face recognition result according to the cosine similarity. According to the invention, a face recognition model based on centralized coordination learning is adopted to carry out feature extraction on the face image, each feature is pulled to an original point and is respectively put into all quadrants, the inter-class distance is larger, the classification efficiency and recognition accuracy of the face are improved, and the method can be widely applied to the technical fieldof face recognition.

Description

technical field [0001] The present invention relates to the technical field of face recognition, in particular to a face recognition method, system and device based on centralized coordination learning. Background technique [0002] Face analysis has always been a hot spot in the field of computer vision due to its important theoretical significance and huge practical application value. How to obtain effective face representations from images (or videos) has always been a core issue in face analysis. Due to the rapid development of deep neural networks and the emergence of large face datasets, face recognition has made great progress in recent years. Face recognition algorithms based on convolutional neural networks emerge in endlessly, from DeepFace, the foundational work of Facebook in 2014, to Google's Facenet in 2015, to SphereFace in 2017, and finally to CosFace and InsightFace in 2018. According to the focus of the loss function, face recognition algorithms can be ro...

AI Technical Summary

Problems solved by technology

[0003] Most of the existing technologies use simple neural networks with no more than 20 layers, such as 8-layer Alexnet, 16-layer or 19-layer VGGnet, or because the applied neural network is too simple, such as DeepFace and 1st, 2nd, 3rd generation DeepId, etc. , the simple neural network can only extract the shallow features of the face, but cannot extract the most representative deep features, which makes the face recognition effect poor

And because the loss function of face recognition is too focused on one of the three categories, the three types of concerns cannot be combined organically. Face features and weight vectors affect each other. Many research works only focus on the expression of feature x or the weight vector. The expression of w ignores the influence of the distribution of face features on the weight vector and the resulting final convergence effect

If most of the face features in a training phase are in the same quadrant, the final features and corresponding vectors will be gathered in this quadrant, which reduces the classification efficiency of different faces.

Images