Comparative Study of the Different Object Detection Algorithms: YOLOv4, SSD, and RCNN based on Accuracy and Speed

摘要

Computer Vision is an interesting and active area of research from past couple of decades. Object detection is a computer vision technique for locating instances of objects in images or videos. It is extensively utilized in various fields such as Self driving, Security surveillance, health care, robotics and so on. The major motivation of this research paper is to test the accuracy and performance of the various popular machine learning algorithms on simple objects under different conditions. Creation of sample images is done using some images found on the internet and some captured at home. Analysis is done for various situations like blurred images, shadows and reflections. This paper explores the features of single-shot detectors (YOLOv4 and SSD) and two-shot detectors (R-CNN) based on speed and accuracy.