Few types of picture samples generation method and device, calculating device and storage medium

Abstract

The present invention is suitable for the computer technology field, and provides a few types of picture samples generation method and device, a calculating device and a storage medium. The method comprises the steps of when a user request of generating the few types of picture samples is received, training a pre-constructed generative adversarial net via the random noise vectors according with the preset distribution and a preset training sample set, according to the trained generative adversarial net, constructing a conditional generative adversarial net, training the conditional generative adversarial net via the random noise vectors, a preset label and the training sample set, generating the few types of picture samples according to the random noise vectors and the few types of labels included in the user request and by the trained conditional generative adversarial net, thereby improving the generalization effect and the quality of the generated few types of picture samples, and further improving the generation efficiency of the few types of picture samples.

Description

Technical field [0001] The invention belongs to the field of computer technology, and in particular relates to a method, device, computing device and storage medium for generating a minority of picture samples. Background technique [0002] With the improvement of artificial intelligence technology, machine learning is increasingly used in people's daily life. Supervised machine learning has become one of the mainstream algorithms in machine learning due to its high accuracy. Supervised learning requires labeled data as the guidance basis for training, and the quality of the training set directly affects the model effect. However, in real life, many available training sets have the problem of label imbalance (label imbalance / class imbalance). For example, using machine learning to determine whether a group of lung images has lung disease, the training set required is the lung images that have been artificially determined with lung disease and the lung images without lung disease...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a method, device, computing device and storage medium for generating minority class picture samples, aiming to solve the lack of universality of the generated minority class picture samples due to the poor noise resistance and generalization effect of the prior art. Sexuality, low generation efficiency, and poor user experience

Images