Survey on Faster Region Convolution Neural Network for Object Detection

Abstract

Convolution Neural Networks uses the concepts of deep learning and becomes the golden standard for image classification. This algorithm was implemented even in complicated sights with multiple overlapping objects, different backgrounds and it also successfully identified and classified objects along with their boundaries, differences and relations to one another. Then comes Region-based Convolutional Neural Networks(R-CNN)which is further more described into two types that is Fast R-CNN and Faster R-CNN. This R-CNN method is to use selective search to extract only 2000 regions from the image and cannot be implemented in real time as it would take 47 sec approximately for each test image. Then comes the fast R-CNN in which changes are made to overcome the drawbacks in R-CNN algorithm in which the 2000 region proposals are not fed to the CNN instead the image is fed directly to the CNN to generate Convolutional feature map. This was then replaced by faster R-CNN which came up with an object detection algorithm that eliminates the selective search algorithm to perform the operation. This algorithm takes 0.2 sec approximately for the test image and we will be using this for real time object detection.So, basically in this paper we are doing research on Faster R-CNN that is being used for object detection method.