Abstract
Object detection as a core role of computer vision, ideals with detecting instances of semantic objects of a certain class in digital images and videos, involving not only recognizing and classifying every object, but localizing each one by drawing the appropriate bounding box around it. With the development of deep Learning, which has recently shown outstanding performance on many fields, many successful approaches to object detection are proposed, making the object detection systems faster and more accurate. In this paper, I review some basic knowledges of object detection, such as the datasets using by most of the papers, the criteria for determing performance, non-maximal suppression for fixing multiple detections. And review the main successful methods in the recent years: R-CNN, Fast R-CNN, Faster R-CNN, R-FCN, YOLO and SSD. By the end of this review, we will hopefully have gained an understanding of how deep learning is applied to object detection, and how these object detection models inspire and diverge from others.Keywords: Object Detection, Deep Learning, CNN, SSD, YOLO
The full paper is here.
