OBJECT DETECTION USING YOLO ALGORITHM

Abstract

Object detection is a common and challenging problem in computer vision.Using underlying deep models, researchers have experimented extensively and contributed to the performance increase of object recognition and related tasks including object classification, localization, and segmentation during the past ten years, thanks to the rapid evolution of deep learning.Object detectors can be broadly divided into two categories: single stage and two stage detectors.Single stage detectors, concentrate on all spatial region proposals for the potential detection of objects via comparatively simpler architecture in a single shot, whereas two stage detectors primarily focus on selected region proposals approach via sophisticated design.The accuracy of the detection and the inference time are the key performance metrics for any object detector.Two-stage object detectors typically have better detection accuracy than single-stage detectors.In contrast to its predecessors, single-stage detectors have a faster inference time.Furthermore, the accuracy of detection is increasing dramatically with the introduction of YOLO (You Only Look Once) and its architectural offspring, often surpassing that of two stage detectors.Instead, then taking detection accuracy into consideration, YOLOs are mostly used in a variety of applications because of their quicker inferences.For instance, the detection accuracies of YOLO and Fast-RCNN are 63.4 and 70, respectively; however, the inference time for YOLO is over 300 times faster.This study presents a thorough analysis of single stage object detectors, with a focus on Yolo objects, including their regression formulation, architecture improvements, and performance statistics.Additionally, we provide an overview of the comparison between two-stage and single-stage object detectors, as well as several YOLO variations, applications utilizing two-stage detectors, and future research possibilities.