Human face recognition method based on dual data enhancement

Abstract

The invention discloses a human face recognition method based on dual data enhancement. A human face data set obtained through the dual data enhancement improves the accuracy of human face recognition. The method comprises the steps that S1, the data set is selected; S2, the data set is preprocessed; S3, primary data enhancement is performed: a generative adversarial network model based on information maximization is built, and training is completed; S4, secondary data enhancement is performed: human face samples generated in the step S3 are subjected to translation, rotation, turnover and zoom processing; and S5, human face classification is performed: the human face samples generated in the step S4 are trained and recognized by selecting a convolutional neural network model.

Description

technical field [0001] The invention belongs to the field of machine learning, in particular to a face recognition method based on double data enhancement. Background technique [0002] Supervised learning is a commonly used method in the field of machine learning. It refers to taking the data in a large-scale data set and the corresponding labels as the input of a certain mathematical model, and then training the model to learn the known data. Features, so as to process the unknown data. According to the definition of supervised learning, the data set is one of the key links to determine the performance of the final model. With the continuous development of deep learning technology and the emergence of open source datasets in different fields, supervised learning methods have been more widely used. However, there are still some deficiencies in the current open source datasets. Taking the face data set as an example, on the one hand, there is a problem of limited overall ...

AI Technical Summary

Problems solved by technology

However, there are still some deficiencies in the current open source datasets.

Taking the face data set as an example, on the one hand, there is a problem of limited overall samples, on the other hand, for a single face sample, it contains very limited feature attributes, and at the same time, the collection, organization, and labeling of the data set is a very time-consuming process. With the process of energy, it is very difficult to include a series of continuously changing feature attributes of all individual face samples in a large-scale data set

Therefore, these problems existing in the current face datasets limit the promotion of face recognition based on supervised learning methods in more application scenarios to a certain extent.

Images