Image semantic segmentation method and device and computer device

Abstract

The present application relates to an image semantic segmentation method and device, and a computer device. The method comprises the following steps: the image to be processed is subjected to convolution processing to obtain a multi-scale feature atlas, and the multi-scale feature atlas is used as the input feature atlas of context interleaving processing; Context interleaving processing is performed on each adjacent feature pair in the input feature set to obtain an interleaved feature set. The interleaved feature atlas is used as the input feature atlas of the context interleaving process, and the adjacent feature atlas of the input feature atlas are returned for context interleaving processing respectively to obtain the interleaved feature atlas until only one interleaved feature atlasis included in the obtained interleaved feature atlas; and the interleaved feature atlas is used as the input feature atlas of the context interleaving process. The interleaved feature map is semantically predicted to obtain a semantic segmented image corresponding to the image to be processed. The contextual information of adjacent feature maps is learned by context interleaving, so that the final interleaved feature maps have better classification characteristics, and then more accurate semantic segmentation images are obtained.

Description

technical field [0001] The present application relates to the technical field of image segmentation, in particular to an image semantic segmentation method, device and computer equipment. Background technique [0002] Image semantic segmentation is one of the important research topics in the field of computer vision and pattern recognition. It is widely used in automatic driving systems, drones, medical imaging and other scenarios. Its goal is to classify each pixel of the image and divide the image into one Group regional blocks with certain semantic meanings, identify the category of each regional block, and finally obtain an image with semantic annotations. Taking the application in the automatic driving system as an example, different types of objects such as people, vehicles, and trees can be segmented and classified through image semantic segmentation, and different annotation methods are used for different types of objects to obtain semantically segmented images, so a...

AI Technical Summary

Problems solved by technology

However, the low-resolution feature map will miss information during convolution, so the feature map obtained by the combination of the above methods also has the problem of information attenuation, which in turn affects the accuracy of semantic segmentation.

Images