Object detection and identification in multiple image scenes using deep learning

Abstract

Object detection is the process of locating objects of interest within an image or video frame. Object detection and classification is a fast growing and an important aspect of research in computer vision. It has yielded variety of applications in shopping systems, health systems, security systems and many surveillance systems. This paper presents a general trainable framework for object detection in images for multiple scenarios. The detection technique will be based on YOLO v4, and Matlab/Simulink software has been used to design and simulate the object detection system. The research work carried out has been able to apply the machine learning technique and also the YOLO image detection technique for the detection of multiple images in different scenes. The study was able to explain the score and recall of the detector depending on the seven features annotated (chair, fire extinguisher, exit sign, clock, printer, screen and trashbin) during the machine learning using the YOLOv4 deep learning object detector. The result of this study revealed that the detector in score plot performed poorly on three classes (printer, screen, and trashbin) but performed well in four classes (chair, clock, exit sign and fire extinguisher). Also, the result of recall plot showed that the printer, screen and trashbin have lower values of 0.3, 0.5 and 0.6 respectively whereas chair, fire extinguisher, exit sign and clock recorded the highest recall value of 1 each. It also suggests that this architecture can be further developed and used in face detection, face recognition, anomaly detection, crowd counting, security surveillance, etc.