Real Time Object Detection Using Deep Learning

Abstract

Real-time object detection remains a critical challenge in numerous domains, holding back advancements in fields like autonomous vehicles, smart surveillance, and beyond. This work directly addresses this hurdle by proposing a high-performance computer vision, real-time system for object detection and classification using a custom-tuned YOLO model with CUDA acceleration. Outperforming prior models, our system achieves a 51.4% mean average precision (mAP) and detects 80 distinct object classes with superior accuracy. Leveraging data from an IoT device, it boasts ~94% in both speed and accuracy. This technology has the potential to revolutionize automated tasks like attendance tracking and industrial quality control, while also paving the way for enhanced object detection in areas like autonomous driving.