Abstract
From the autonomous car driving to medical diagnosis, the requirement of thetask of image segmentation is everywhere. Segmentation of an image is one ofthe indispensable tasks in computer vision. This task is comparativelycomplicated than other vision tasks as it needs low level spatial information.Basically, image segmentation can be of two types: semantic segmentation andinstance segmentation. The combined version of these two basic tasks is knownas panoptic segmentation. In the recent era, the success of deep convolutionalneural network (CNN) has influenced the field of segmentation greatly and gaveus various successful models till date. In this survey, we are going to take aglance on the evolution of both semantic and instance segmentation work basedon CNN. We have also specified architectural details of some state-of-the-artmodels and discuss their comparative training details to present a lucidunderstanding of hyper-parameter tuning of those models. We have also drawn acomparison among the performance of those models on different datasets.