A Face Recognition Method Based on Deep Learning Model and Migration Learning

Abstract

The invention discloses a face recognition method based on a deep learning model and transfer learning, comprising the following steps: preprocessing source images and target images and setting corresponding labels, the number of source images is M, and the number of target images is N, M >N; establish a source neural network whose classifier output dimension is M; construct a source data set based on source image features and labels and use the source data set to train the source neural network, optimize the model parameters through the neural network BP algorithm, and obtain the source training model; Establish the target neural network with the output dimension of the classifier as N and initialize the target neural network with the parameters of the source training model; construct the target data set based on the target image features and labels and use the target data set to train the target neural network, and update it by dynamically selecting K The algorithm performs gradient descent to optimize model parameters to obtain a target training model; image recognition is performed through the target training model; the invention improves the accuracy and robustness of the face recognition model.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a face recognition method based on a deep learning model and transfer learning. Background technique [0002] Face recognition is a biometric technology for identification based on human facial feature information. The current methods mainly focus on the following aspects: [0003] (1) Template matching, there are mainly two methods, fixed template and deformed template; the process of fixed template method first uses an algorithm to obtain one or several reference feature templates of the target, and then uses a certain measure to calculate the test sample The similarity between the feature template and the reference template judges whether the test sample is the target face by whether the result is greater than the threshold; It is difficult to obtain an effective parameter feature template to represent the commonality of faces; the deformation template is to improve...

AI Technical Summary

Problems solved by technology

[0007] However, the face recognition system is dependent on the specific application. The face image is affected by many factors such as ambient light, angle of view, expression, makeup, etc., resulting in different face databases used in different application backgrounds.

A high-precision face recognition system often needs to use a large number of face samples to learn the face recognizer. If the current application background, the limited number of face database samples will inevitably affect the accuracy of face recognition

Images