Advancing Surveillance Technology: YOLO-Based Machine Learning Approach for Shadow Detection

Surveillance cameras are essential for security, but sometimes capture objects is overlooked, especially in challenging lighting conditions. This paper presents a real-time shadow detection and classification model based on YOLO. It can recognize the shadows of cows, dogs, guns, and people, improving security by exposing hidden dangers. The importance of precise shadow detection is emphasized in the paper, which also offers insights into the model’s technical details, dataset creation, and practical applications. The study also examines convolutional neural networks (CNNs) for real-time processing and describes the YOLOv8 architecture. The model’s remarkable mean average precision of 0.817, attained with the carefully annotated custom dataset, emphasizes its applicability in various scenarios, and it has the potential to improve security cameras.