WatchDog- A Smart Surveillance System

Smart Surveillance Systems have become increasingly important in modern urban environments where public safety, automated monitoring, and rapid threat detection are critical. Traditional CCTV-based surveillance suffers from significant limitations, including dependency on human operators, fatigue, slow response times, and inability to detect complex real-time events. This research introduces an AI-driven Smart Surveillance System that leverages deep learning—particularly Convolutional Neural Networks (CNNs) and YOLO-based object detection—to automatically identify and classify critical events from video feeds. The system processes live CCTV footage, extracts spatial-temporal features, and detects activities such as accidents, falls, crowd formation, weapon visibility, and suspicious human behavior. By converting video frames into structured data representations and applying CNN-based detection pipelines, the system captures detailed visual patterns and contextual cues, leading to high accuracy and robustness. The model was trained and evaluated on standard datasets from COCO, Open Images, UR Fall Detection, and custom surveillance footage to achieve reliable real-time detection performance. This study demonstrates the effectiveness of deep learning–based architectures in building scalable, intelligent, and responsive surveillance solutions suitable for smart cities and public security infrastructures.