Benchmarking Reidentification in Multi-Camera Tracking Systems with YOLOv8 and ResNet-50

The primary aim of this paper is to benchmark reidentification within a multi-camera tracking system. This benchmark has been developed by leveraging transfer learning, utilizing YOLOv8 for real-time object detection and ResNet-50 for feature extraction. The objective is to evaluate the system’s performance in accurately reidentifying vehicles across multiple cameras in real-world traffic surveillance scenarios. This benchmarking endeavor aims to provide a standardized evaluation framework for assessing the capabilities and limitations of vehicle reidentification techniques, with a focus on their applicability in challenging conditions such as low-light environments, image compression, and object occlusions.