A Deep Learning Approach for Object Detection and Instance Segmentation using Mask RCNN

Maibam Mangalleibi Chanu, Ravi Lourembam and Dr. Khelchandra Thongam

The computer vision has rapidly upgraded in area of object detection, localization and segmentation over short period of time. How fast and consistent has become a key question in this vision community. With great increased in techniques, the accuracy for object detection has improved significantly. While performing such enormous task a systematic search is needed so as not to miss out any object as far as possible. This however required extensive efforts and time. In this paper we show that Mask R-CNN can be deployed in this field. Mask RCNN is a recently propose state-of-art algorithm. It is a simple and wide-ranging framework for bounding-box object detection and segmentation of images with a mask. Mask R-CNN is an extension of Fast R-CNN by adding a prediction of each objects in an image along with the existing method of bounding box recognition. The proposed method shows higher accuracy in detecting objects.

Volume 12 | 03-Special Issue

Pages: 95-104

DOI: 10.5373/JARDCS/V12SP3/20201242