Instance Segmentation using Mask RCNN for Surveillance

Abstract

Instance Segmentation combines object detection, where the goal is to classify and locate every object according to its bounding box, and semantic segmentation, which defines each pixel according to its object category. Deep learning uses Mask R-CNN architecture to perform instance segmentation on data. This architecture is an enhancement of a well-known object detection algorithm called Faster R-CNN. With the addition of a mask branch to the Faster R-CNN model, the Mask R-CNN model has been evolved. This paper contains the idea of how Mask R-CNN can perform instance segmentation by using some examples of surveillance detection and segmentation. In this algorithm, masks are applied to the objects present in the image in order to detect and classify them. Mask RCNN algorithm is more accurate and has a lower time complexity than existing Faster RCNN algorithm for video surveillance.