Performance Analysis of Object Detection Algorithms for Multi-Person Identification
Vasantha Sandhya Venu, Pooja Cherupally, Anji Yadav Kaitha, Aishwarya Belly, Leema Nelson, Shanmugasundaram Hariharan · 2023
In the prospect of computer vision, the accurate recognition of individuals and objects constitutes a pivotal pursuit with far-reaching implications. It has enormous significance for a variety of uses, including retail analytics, public transportation, event management, the hospitality industry, workplace utilization, education facilities, public safety and many more. This paper delves into the intricacies of object detection algorithms for multi-person identification. It involves detecting and recognizing individuals within images containing multiple persons. Object detection algorithms locate and draw bounding boxes around objects of interest in an image in this case it is people. We have implemented HOG-SVM, YOLOv8 and RetinaN et algorithms to count people in varied-sized gatherings in different contexts with multiple postures. The performance of these algorithms is analyzed, where recall of 0.97, 0.94, and 1.0 are observed for HOG-SVM, YOLOv8 and RetinaN et algorithms respectively demonstrating their ability as people counting algorithms.