Image segmentation model with coding and decoding structure combined with attention mechanism and training method thereof

Abstract

The invention discloses an image segmentation model with a coding and decoding structure combined with an attention mechanism. The image segmentation model comprises a coding module, a decoding module, a coding and decoding connection module and a pixel classification layer, the coding module comprises a plurality of coding units which are connected in series, the decoding module comprises a plurality of decoding units which are in one-to-one correspondence with the coding units, the coding and decoding connection module is used for connecting the coding module and the decoding module, and the pixel classification layer is used for independently generating a category probability for each pixel to obtain an image segmentation result. The image segmentation model combines the modeling advantage of SegNet in local environment information and the advantage of Transform in learning global semantic association, the limitation that SegNet is easy to over-fit when a data set is small is improved by enhancing global semantic association, and the network segmentation performance is improved at the same time.

Description

technical field [0001] The invention relates to the field of visual image processing, in particular to an image segmentation model with a codec structure combined with an attention mechanism and a training method thereof. Background technique [0002] Image segmentation is an important part of image recognition and computer vision, and has broad application scenarios in various fields. In the field of computer vision, image segmentation technology can be classified into two categories after a long period of development: one is the traditional image segmentation method based on artificially extracted features, and the other is the image segmentation method based on deep learning to extract features. [0003] Image segmentation methods based on artificially extracted features, such as threshold-based image segmentation methods, edge detection-based image segmentation methods, region-based image segmentation methods, etc., but traditional algorithms are only for specific image ...

AI Technical Summary

Problems solved by technology

[0006] The existing SegNet network can make full use of the local information of the image, has high segmentation accuracy and training efficiency, and is widely used in multi-class segmentation tasks, but usually requires large training data, otherwise it is easy to overfit

Images