Network training method and device, image recognition method and electronic equipment

Abstract

The invention relates to a network training method and device, an image recognition method and electronic equipment. The method comprises steps of: according to a first image group in a training set,training an image identification network in an initial state to acquire an image identification network in a first state, wherein the first image group at least comprises labeled first sample images;training the image recognition network in the first state according to a second image group in the training set to obtain an image recognition network in a second state, the second image group comprising labeled first and second sample images and an unlabeled third sample image; and training the image recognition network in the second state according to the first image group to obtain an image recognition network in a target state. According to the embodiment of the invention, the method can achieve the full utilization of the rough labeling and non-labeling image data, reduces the data demands for fine labeling, improves the training effect of the recognition network, and improves the segmentation and recognition precision of the image recognition network for the image data.

Description

technical field [0001] The present disclosure relates to the technical field of computer vision, in particular to a network training method and device, an image recognition method and electronic equipment. Background technique [0002] Deep learning has achieved great success in the fields of natural image processing and recognition, and has also made great progress in traditional image processing algorithms in medical imaging. Using deep learning technology to automatically segment, extract and identify types of tissues, organs and lesions from medical imaging data is of great reference significance for assisting doctors in disease diagnosis and improving their work efficiency. [0003] Existing medical image segmentation and recognition algorithms are fully supervised deep learning models. In the model training phase, it is necessary to fully label each slice of each training medical image data, that is, it is necessary to accurately delineate the target area and its attr...

AI Technical Summary

Problems solved by technology

This image segmentation and recognition algorithm based on fully supervised deep learning relies on a large amount of finely labeled data to obtain better segmentation and recognition results. When the finely labeled data is insufficient, the performance of fully supervised deep learning algorithms is greatly restricted. The accuracy of the trained model cannot accurately segment and identify medical images

Images