The Application of Attention Mechanism in Semantic Image Segmentation

Abstract

Image semantic segmentation technology has been greatly developed, and the application of attention mechanism to semantic segmentation has become a new research and development direction. We investigated the application of attention mechanisms in image semantic segmentation in recent years. According to its subdivision, it can be divided into a method based on attention region segmentation and a method based on pixel attention segmentation. The method based on pixel segmentation can also be subdivided into full-supervised attention mechanism segmentation and weak supervisory attention segmentation. We also summarize the semantic segmentation datasets in recent years and summarize the frequency of use for readers. Finally, we summarize the general direction of the semantic segmentation of attention mechanisms and future trends.