Semi-supervised semantic segmentation model training method, semi-supervised semantic segmentation model identification method and semi-supervised semantic segmentation model identification device

Abstract

The embodiment of the invention provides a semi-supervised semantic segmentation model training method, a semi-supervised semantic segmentation model identification method and a semi-supervised semantic segmentation model identification device. The semi-supervised semantic segmentation model training method of the embodiment comprising obtaining first supervised data obtained by manually annotating a to-be-annotated object in a first image, and then obtaining a full-supervised semantic segmentation model with a relatively high recognition rate of the to-be-annotated object through training ofthe first supervised data; labeling the to-be-labeled object in the second image which is not manually labeled by utilizing the fully-supervised semantic segmentation model to obtain second superviseddata; training a semi-supervised semantic segmentation model by utilizing the first supervised data obtained through manual annotation and the second supervised data obtained through full-supervisedsemantic segmentation model annotation, and identifying the first image, the second image and the random disturbance item by utilizing the semi-supervised semantic segmentation model to obtain third supervised data; and finally, training the semi-supervised semantic segmentation model again through the first supervised data, the second supervised data and the third supervised data.

Description

technical field [0001] One or more embodiments of this specification relate to the technical field of image processing, in particular to a semi-supervised semantic segmentation model training method, recognition method and device. Background technique [0002] Image recognition refers to the technology of using computers to analyze and understand images to identify targets and objects in various patterns. Image recognition technology has been used in remote sensing image recognition, communication, image file restoration and other fields to provide convenience for people's lives. [0003] At present, the models used for image recognition are usually obtained after training with training data, and the training data is obtained through manual processing, which leads to insufficient generalization ability of the model and affects the accuracy of image recognition. Therefore, to address the above deficiencies, it is necessary to provide an image recognition model with higher ac...

AI Technical Summary

Problems solved by technology

[0003] At present, the model used for image recognition is usually obtained after training with training data, and the training data is obtained through manual processing, which leads to insufficient generalization ability of the model and affects the accuracy of image recognition

Images