Analysis on Approaches and Structures of Image Semantic Segmentation

Abstract

In recent years, due to the increasing demand for the understanding and recognition of content in images, image semantic segmentation technology has developed rapidly. Image semantic segmentation technology has also seen more and more reforms and innovations Each classical model has its own innovation and characteristics, which contributes to the development of image semantic segmentation.In this paper, four popular semantic segmentation models are reviewed and their characteristics are introduced.The results show that compared with other models, the SETR model based on Transformer has a higher performance level in semantic segmentation results.