Optimizing Object Detection With YOLOv5 fOR Real-Time Applications

Object detection has emerged as one of the most important tasks in computer vision that enables machines to identify and localise objects within images or videos. Recent advances in object detection, such as YOLOv11 and transformer-based architectures like DETR, have greatly improved the accuracy, speed, and efficiency of detection. This paper follows the path of the evolution of object detection models, with special attention to innovations such as nonmaximum suppression-free training, dynamic head design, and lightweight architectures. We test these models across real-time applications, including autonomous vehicles, healthcare imaging, and robotics. Additionally, we address challenges of overlapping objects, computational efficiency, and optimisation for edge devices. Using empirical analysis, we demonstrate the trade-off between speed and precision, highlighting what this means for the future of AI-driven object detection technologies. KEYWORDS–Object detection, YOLOv11, DETR (DEtection TRansformer), lightweight architecture, objects autonomous vehicles, computational efficiency.