The Journal of Science and Engineering Applications (JSEA)

Abstract

From the first YOLO version to the last version, it was considered the most common, acceptance and speed object detection algorithm. YOLO is a single-step object detection algorithm since it performs both detection and classification at the same step. This paper focuses on and highlights on the key contributions of each version and the modifications that happened in each one. Most modifications tend to improve either the accuracy or speed of the model or to improve both of them for example by manipulation of the number of layers or by the innovation of new technologies such as Bag of Freebies or to innovation a new type of backbones. So, this paper reviews these approaches and how much they effect on the model performance by making a comparison between the different versions of YOLO algorithms.