Multi-scale image translation method based on semi-supervised learning

Abstract

The invention discloses a multi-scale image translation method based on semi-supervised learning, and the method comprises the steps: enabling a multi-scale discrimination generative adversarial network to complete all image translation tasks at the same time through a unified frame, and carrying out the discrimination of a plurality of scales of a generated image at the same time, and preventingmany unreasonable forged objects from locally appearing in the generated image; combining a generative adversarial network and dual learning to train a model by utilizing non-paired data, improving the semi-supervised image translation effect, meanwhile, effectively utilizing non-paired image information to improve the performance of the model, and reducing the requirement of model training for paired images. According to the multi-scale image translation method based on semi-supervised learning, the defect that a supervised image translation algorithm needs a large amount of training data iseffectively overcome, model convergence can be accelerated, and themodel performance is improved. The method has obvious effect and is suitable for wide popularization.

Description

technical field [0001] The present invention relates to the technical field of image translation, in particular to a multi-scale image translation method based on semi-supervised learning. Background technique [0002] Image translation refers to the task of automatically transforming one scene representation of an image into another scene. At present, convolutional neural networks are used as models for image translation, but for model construction of specific tasks, the loss function design and optimization strategies are not the same, which greatly increases the burden of model design. [0003] By minimizing the Euclidean distance between model predictions and ground truth labels the model outputs blurry images. Image translation algorithms based on supervised learning require a large amount of paired input-output training data. Existing image translation tasks are based on supervised learning models, which require a large amount of paired training data. However, in rea...

AI Technical Summary

Problems solved by technology

[0005] In view of the above defects, the technical problem solved by the present invention is to provide a multi-scale image translation method based on semi-supervised learning to solve the complex construction of different task models in the current technology, the burden of model design, and the data required for model training and learning. volume problem

Images