Semantic Segmentation using Deep Learning Approaches - A Study

Semantic segmentation is an important technology which gets commonly used in medical imaging, autonomous driving vehicles, and also backgrounds for virtual meetings. Classes can be different real-world objects such as roads, cars, bicycles, people, trees, lanes, trucks, buildings etc. Classes can correspond to different anatomical structures and organs when considering medical images. Semantic segmentation is a broadly applicable technology. The techniques discovered to improve current semantic segmentation methods could also lend themselves to improving other dense prediction tasks. These tasks could include optical flow prediction (object motion prediction tasks), image super-resolution such as in remote gaming or in video resolution enhancement, and so on. This paper briefly presents a survey on existing work conducted to achieve semantic segmentation of image problems with the use of deep learning methods as well as image processing approaches. Deep learning provides several methods for semantic segmentation such as 2D convolution networks, 3D convolution networks, etc. This paper discusses the classification, challenges, application, and methods for semantic segmentation.