SPARSE R-CNN OBJECT DETECTION USING PROPOSAL BOXES

Abstract

Artificial Intelligence has advanced a lot over the last few years to solve various challenges in the modern day world. One of these challenges is the recognition of images digitally and extracting important features.The topic of deep learning has created an impact in this field and the major algorithm that contributed is convolutional neural network.Convolutional neural network has been continuously improving to get much efficient algorithms to extract the features better.This report will be covering object detection various algorithms and how they work. The methods to detect objects are mainly divided into three groups: sparse methods, dense methods and dense to sparse.The dense methods divides image to grid to slide proposal box to detect the object, while the dense to sparse tries to find important regions where the object will be present using region proposal. Both of these methods are great but to further improve the algorithm , sparse RCNN was introduced which takes away less user input parameters to calculate the location and detect where the object is present in the image. It uses iterative learning to predict where the object will be present in the image with different sizes of proposal boxes.Then it classifies and predicts what object it is thus creating more efficiency.