Object detection through region proposal based techniques

Abstract

Object Detection is the most significant operation to be performed for applications based on Video surveillance, Medical diagnosis, Image processing, Robotics and many more. Recently many researchers are using an Object detection approach which leverages multiple layers of different CNN architectures to detect significant features. The Generic object detection is performed either through region based proposal technique or regression technique. This paper gives the overview of R-CNN, Fast R-CNN and Faster R-CNN which are region proposal-based object detection techniques. In this paper comparison is done on the basis of Mean Average Precision (MAP) score obtained for various region proposal based object detection methods in IMAGENET Large Scale Visual Recognition Challenge (ILSVRC) and PASCAL VOC Challenges. Also comparison is done among ILSVRC and PASCAL VOC challenges with respect to different parameters and significant differences are reported along with critical review.