A Review on Different Deep Learning Approaches For Semantic Segmentation

Abstract

Image semantic segmentation is increasingly being of enthusiasm for computer vision and AI specialists. Numerous applications on the ascent need precise and productive segmentation components: automatic driving, indoor navigation, and augmented reality frameworks to give some examples. This interest corresponds with the ascent of profound learning approaches in pretty much every field or application target identified with PC vision, including semantic division or scene understanding. This paper gives a survey on profound learning strategies for semantic division applied to different application regions. Right off the bat, we depict the phrasing of this field just as obligatory foundation ideas. Next, the primary datasets and difficulties are presented to support specialists choose which are the ones that best suit their needs and their objectives. At that point, existing techniques are investigated, featuring their commitments and their centrality in the field. At long last, quantitative outcomes are given for the portrayed techniques and the datasets in which they were assessed, catching up with a discourse of the outcomes. Finally, we pay attention to many promises for future work and make our own decision related to best techniques for semantic segmentation using deep neural network.