Instance segmentation method and device, electronic equipment and storage medium

Abstract

The invention provides an instance segmentation method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining a category response graph, an instance center offset field and an instance boundary response graph of a to-be-segmented image; establishing a prior template of the target according to the attitude of the target in the to-be-segmented image; based on the category response diagram, the instance center offset field, the instance boundary response diagram and a prior template, respectively improving the category response diagram and the instance boundary response diagram through template matching to obtain a corresponding template-based category response diagram and a template-based boundary diagram; and carrying out region expansion based on the template category response diagram under the limitation of the template boundary diagram to obtain an instance segmentation result. According to the scheme, under-segmentation and over-segmentation can be reduced.

Description

technical field [0001] The present invention belongs to the technical field of image segmentation, and particularly relates to an instance segmentation method, device, electronic device and storage medium. Background technique [0002] Existing instance segmentation techniques in the field of image segmentation can be mainly divided into two categories, one is fully supervised instance segmentation, and the other is weakly supervised instance segmentation. The first category of fully supervised instance segmentation methods, instance segmentation techniques based on deep learning methods can automatically predict the category of each pixel in an image and the instance to which it belongs. However, the main disadvantage of this method is that in the process of training the neural network, accurate image pixel-level category and instance annotations need to be used, and this pixel-level annotation requires a lot of manpower and time investment, and the time and economic costs ...

AI Technical Summary

Problems solved by technology

Taking image category labeling with the lowest labeling cost as an example, the existing weakly supervised instance segmentation methods based on image category labeling mainly perform pixel-level instance segmentation based on perceptual image information, which is prone to produce segmentations that obviously do not conform to human prior cognition result

Images