Attention mechanism-based lightweight semantic segmentation model construction method

Abstract

The invention discloses an attention mechanism-based lightweight semantic segmentation model construction method, which is applied to the technical field of image processing, and a training set is formed by giving an image I and a corresponding real label graph GT. The method comprises the steps of step 1, establishing a model; step 2, model training; and step 3, model testing: inputting a test set image into the trained network model to obtain a test result. According to the invention, the image segmentation accuracy and segmentation speed are improved; the segmentation process is not easy to over-fit; efficiency is high, and actual deployment is facilitated; and under the condition that the annotation data is insufficient, the annotation data is quickly trained, so that the performance is further improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for constructing a lightweight semantic segmentation model based on an attention mechanism. Background technique [0002] Image segmentation refers to the computer vision task of marking the specified area according to the content of the image. Specifically, the purpose of image semantic segmentation is to mark each pixel in the image and associate the pixel with its corresponding category. It has important practical application value in scene understanding, medical image, unmanned driving, etc. [0003] Classic semantic segmentation models include: [0004] Fully convolutional neural network (FCN), as a classic production of semantic segmentation network in deep learning, draws on the traditional classification network structure, but is different from the traditional classification network, and converts the fully connected layer of the traditional classificati...

AI Technical Summary

Problems solved by technology

This ensures the integrity of the high-frequency information, but when unpooling the lower-resolution feature map, the information between the neighbors of the pixels is also ignored.

Images