Video Objects Detection Using Deep Convolutional Neural Networks

Abstract

Deep Convolutional Neural Network (CNN) technology has attracted the interest of researchers in computer vision to discover objects in photos and videos and recognize faces and their characteristics because it has effective performance. This paper describes the main features of CNN's most accurate deep learning and its applications. The YOLOv3 model was used to accurately classify and predict objects in each frame of a video by embedding the object box and extracting features using the ImageAI library. This library supports pre-trained models on the coco dataset to detect pedestrians and objects using training data (video about Kirkuk University). It is possible to use the importance of this study to monitor the traffic or search for something tracked by video surveillance. In addition to detecting pedestrians in the video, visible objects, and faces. The result compares this video detection method and is evaluated using performance metrics such as a map. It achieved the best video processing performance with an accuracy of 99.52% and very high accuracy for video object detection (98.97% people, 99.12% cars, 100% bicycle, 100% clock), which proved to be higher performance compared to other detection models.