Deep Learning Models for Image Segmentation

Abstract

Artificial Intelligence and deep learning models have evolved rapidly in the last decade and successfully applied to face recognition, autonomous driving, satellite imaging, robotics, and many more. Computer vision tasks often require adequate segmentation of an image that helps to understand the patterns and information. The adequate segmentation makes the analysis of each part of an image easier. Traditional segmentation techniques are often applied for image segmentation, but they are less efficient than deep learning techniques. Using deep learning approaches, it is possible to obtain hierarchical feature representations directly from the images, and hence, it eliminates the requirement of handcrafted features. This paper covers the fundamentals of image segmentation and deep learning, deep learning models for image segmentation, some successful implementations of deep learning models for image segmentation, and available open and benchmark datasets for image segmentation tasks.