This research offers a comparative analysis of the YOLOv5 and YOLOv8 designs for person detection utilizing the DukeMTMC-reID dataset. We assess both models based on their performance parameters, encompassing accuracy, speed, and resilience in a re-identification context. Our investigations demonstrate significant disparities in performance and processing efficiency between YOLOv5 and YOLOv8, offering insights into their appropriateness for surveillance systems. The findings demonstrate that YOLOv8 provides enhanced detection accuracy, but with increased computing requirements relative to YOLOv5. The results are to assist researchers and practitioners in choosing the suitable model for video surveillance applications.