Image Semantic Segmentation Model Construction Method and Device Based on Generative Adversarial Network

Abstract

This application discloses a method and device for constructing an image semantic segmentation model based on a generative confrontation network. The method uses a segmentation model to train a source domain dataset, and then uses a generative confrontation network to convert the source domain dataset into a new target domain dataset. This new target domain dataset retains the structural features of the images in the source domain dataset but at the same time has the global features of the target domain dataset, so fine-tuning the source domain segmentation model with the new target domain dataset will reduce the difference between the source domain and the target domain. The displacement effect, and will not have a negative impact on other image features of the data, improves the generalization ability of the image semantic segmentation model, and improves the accuracy and efficiency of the adaptive image semantic segmentation model. In this way, by using the generative confrontation network, the influence of the domain shift between the source domain and the target domain is effectively reduced, the accuracy and efficiency of adaptive image semantic segmentation are improved, and the cost is reduced. the accuracy.

Description

technical field [0001] The present application relates to the technical field of image recognition, in particular to a method and device for constructing an image semantic segmentation model based on a generative confrontation network. Background technique [0002] Image segmentation refers to the computer vision task of marking the specified area according to the content of the image. Specifically, the purpose of image semantic segmentation is to mark each pixel in the image and associate the pixel with its corresponding category. It has important practical application value in scene understanding, medical image, unmanned driving, etc. [0003] The traditional image semantic segmentation methods are as follows: [0004] The first is the threshold method, which converts a grayscale image into a binary image with background separation; [0005] The second is the method of pixel clustering, assuming that there are K categories in the image, and the pixel points in the image ...

AI Technical Summary

Problems solved by technology

However, the success of deep learning largely depends on the quality of the training set used. A high-quality training set requires a lot of manpower and material resources. Although there are currently several high-quality image semantic segmentation The data set can help scholars to carry out related research work, but due to the limitation of the generalization ability of the model itself, it cannot be well applied to the image semantic segmentation in real life, so the adaptive problem of the image semantic segmentation model needs to be solved. hot issues

Images