Instance segmentation method for correcting classification errors by using classification attention module

Abstract

The invention discloses an instance segmentation method for correcting classification errors by using a classification attention module. The method comprises the steps of obtaining a plurality of feature maps of a to-be-processed image based on a backbone neural network of a preset instance segmentation model; performing convolution processing on the feature map based on a classification module ofa preset instance segmentation model to obtain a semantic category of the to-be-processed image; performing convolution processing on the feature map based on a classification attention module of a preset instance segmentation model to obtain a pixel category of the to-be-processed image; determining a foreground class channel of the to-be-processed image based on the pixel foreground class channel and the semantic class; and performing convolution processing on the mask convolution kernel parameter of the foreground class channel of the to-be-processed image and the mask feature map so as toobtain the prediction mask of the foreground class channel of the to-be-processed image, instance segmentation is performed on the image according to the prediction mask, and therefore, misclassification instances can be corrected, and the accuracy of image instance segmentation is further improved.

Description

technical field [0001] The invention relates to the technical field of image segmentation, in particular to an instance segmentation method using a classification attention module to correct classification errors. Background technique [0002] With the development of deep neural network technology, various image instance segmentation models have emerged. The algorithm with the best performance indicators in the current academic community is the SOLO algorithm (Segmenting Objects by Locations) and its improved version SOLOv2. SOLOv2 has made two improvements to the detection effect and operating efficiency of the mask: (1) mask learning: it can better learn the mask; (2) mask NMS: the matrix nms is proposed, which greatly reduces the time of forward reasoning. Compared with SOLOv1, SOLOv2 has improved average accuracy and speed. [0003] However, in the existing image instance segmentation methods, the category of the segmented image is determined by semantic classification....

AI Technical Summary

Problems solved by technology

[0005] The present invention provides an instance segmentation method using a classification attention module to correct classification errors, to solve the technical problem that the misclassified instances cannot be corrected in the prior art, and the accuracy of image instance segmentation is not high. The method includes:

Images