Image segmentation method, system, medium and electronic terminal

Abstract

The present invention provides an image segmentation method, system, medium and electronic terminal. The method includes: obtaining training images to be segmented; inputting pairs of training images to be segmented into a positioning network to obtain a positioning feature map, wherein the step of obtaining the positioning feature map includes : Extract the same point feature map and the overall attention feature map for the paired training images to be segmented, and obtain the positioning feature map according to the overall attention feature map; input the positioning feature map into the segmentation network for training, and obtain the image segmentation model; The paired tumor images to be segmented are input into the image segmentation model to perform tumor image segmentation; the image segmentation method in the present invention not only extracts the same semantic features, but also extracts different semantic features by inputting the paired training images to be segmented into the positioning network. feature, and input the obtained positioning feature map into the segmentation network for training, obtain the image segmentation model, realize the semantic extraction across images, and achieve high segmentation accuracy.

Description

technical field [0001] The present invention relates to the technical field of image segmentation, in particular to an image segmentation method, system, medium and electronic terminal. Background technique [0002] With the development of image segmentation technology, weakly supervised image segmentation learning has received more and more attention. Weakly supervised image segmentation learning uses image-level annotations with few annotations instead of pixel-by-pixel annotations. For example, weakly supervised images for tumor images Segmentation, the existing weakly supervised image segmentation learning usually obtains the location feature map through the classification network to supervise the segmentation network learning. However, at present, in the process of weakly supervised image segmentation learning, only the same location feature map is retained Semantic features, and only pay attention to the information in the current image, can not obtain more complete ob...

AI Technical Summary

Problems solved by technology

[0003] The present invention provides an image segmentation method, system, medium, and electronic terminal to solve the problem of generating strategies for positioning feature maps in the prior art, only retaining the same semantic features, and only paying attention to the information in the current image, and cannot obtain better results. Complete object information, resulting in poor image segmentation and low segmentation accuracy

Images