Image instance segmentation method, device and apparatus based on feature fusion and storage medium

Abstract

The invention provides an image instance segmentation method, device and apparatus based on feature fusion and a storage medium, and the method comprises the steps: obtaining a to-be-segmented image,extracting features of the to-be-segmented image through a ResNet convolution unit with a BN layer removed to acquire a feature map; inputting the feature map and an original feature map into a feature fusion unit for addition and fusion, inputting the fused feature map into a 3 * 3 convolution unit, readjusting the feature map, extracting background context information from the adjusted feature map through a pooling chain, and fusing a plurality of feature maps of different receptive fields to obtain a pooled feature map; inputting the pooled feature map into a convolution unit, and integrating the result of the pooling chain to obtain an integrated feature map; and obtaining an instance segmentation result through classification segmentation according to the integrated feature map. According to the invention, processing is carried out through a self-defined feature fusion module, and feature map information is utilized more efficiently, so that finally processed mask map informationis richer and higher in accuracy.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, in particular to a method, device, equipment and storage medium for image instance segmentation based on feature fusion. Background technique [0002] The image instance segmentation problem is an extremely important sub-problem in the field of computer vision. Its main purpose is to accurately intercept the objects of interest in the image from the image, so as to avoid background interference and facilitate subsequent processing. Instance segmentation itself contains many branch problems, namely object detection and semantic segmentation. Every point of improvement is extremely important to the improvement of the final result, so the improvement direction of this article can also start from these two places. [0003] Early image segmentation did not distinguish between instance segmentation and semantic segmentation, the purpose is to distinguish the foreground and background ima...

AI Technical Summary

Problems solved by technology

Early SDS used MCG to generate proposal areas, and then used Bottom-Up to segment the results, but this was a large amount of calculation and the segmentation effect was average.

Images