GAN-based enhanced sample generation method

Abstract

The invention relates to a GAN-based enhanced sample generation method. The method comprises the following steps: acquiring a to-be-strengthened sample; inputting the to-be-strengthened sample into a pre-trained StrGAN generation model for strengthening to obtain a target strengthened sample; the training mode of the StrGAN generation model is as follows: obtaining an original picture sample; processing the original picture sample by adopting a method of converting RGB into YCbCr to obtain a processed picture sample; inputting the processed picture sample into a generator based on a StrGAN algorithm, and automatically extracting the characteristics of the input processed picture sample by adopting the StrGAN algorithm to generate a reinforced sample; performing precision evaluation on the enhanced sample, and when the precision reaches a preset condition, obtaining the StrGAN generation model, so that the generated target enhanced sample improves the neural network from the positive influence of disturbance, and the performance of the neural network is improved.

Description

technical field [0001] This application relates to the technical field of computer image processing, in particular to a method for generating enhanced samples based on GAN. Background technique [0002] With the development of artificial intelligence technology, machine learning and deep learning algorithms have been widely used in many complex fields, such as object detection, face recognition, natural language processing and image classification, etc. However, some studies have found that neural networks are susceptible to small input disturbances, and most current research (such as adversarial examples) uses the negative effects of disturbances to make neural networks misjudge. [0003] At present, the generation methods of adversarial examples are mainly divided into traditional methods and generative adversarial network methods. Among them, traditional methods are mainly based on gradient and optimization methods, such as FGSM (Goodfellow I J, Shlens J, Szegedy C. Expl...

AI Technical Summary

Problems solved by technology

[0004] The above-mentioned generation methods of adversarial samples are all designed based on the idea that the negative impact of disturbance will mislead the network. The neural network is easily affected by small disturbances, so it is easy to make the neural network misjudgment, making the performance of the neural network low.

Images