Method for data augmentation, apparatus and system and computer readable storage medium

Abstract

The embodiments of the invention disclose a method for data augmentation. The method comprises the following steps that: original image information is received, wherein the original image informationis training samples in a data set; a plurality of feature images of an original image are obtained, wherein each feature image contains the depth feature information of the original image; each feature image is expanded, so that a plurality of expanded images of each feature image can be obtained; and all the expanded images of the plurality of feature images are added into the data set as training samples. The embodiments of the invention provide a depth feature information-based data augmentation method. According to the depth feature information-based data augmentation method, data augmentation is carried out on the basis of the depth feature information of an image, and therefore, the essence information of an object can be found from the depth features of the object. With a depth feature information-based data augmentation technology and a face recognition network combined, the training of a recognition classifier can be realized more effectively. The embodiments of the inventionalso disclose an apparatus for data augmentation, a system for data augmentation and a computer readable storage medium.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a method, device and system for data augmentation, and a computer-readable storage medium. Background technique [0002] Face recognition technology is an important link in computer vision applications, and has broad application prospects in many practical application scenarios. Face recognition technology can be divided into two main tasks: face verification technology and face classification technology. The verification technology needs to input two faces, and determine whether the two faces are the same person through calculation. There is a certain difference between the classification technology and the verification technology. It first extracts the features of a picture, and then compares the features of the picture with all the features of the pictures in the database to select which person in the database the picture is. With the continuous devel...

AI Technical Summary

Problems solved by technology

There are some traditional amplification methods that have changed a lot on the basis of the original image, but the changes to the specific core depth feature information are relatively small, so that the real purpose of data amplification cannot be achieved; while other methods have changed the depth feature information too much , resulting in a long distance between the original data and the modified data, which makes training difficult and reduces training efficiency and convergence time

Images