An End-to-End Real-Time YOLO-Based Object Detection and Alert Framework with Web Integration

Real-time object detection plays a critical role in intelligent surveillance and safety monitoring systems where rapid identification of obstacles and timely alert generation are essential for improving situational awareness. This paper presents an end-to-end real-time object detection and alert framework that integrates a lightweight YOLO-based deep learning model with a web-based streaming architecture. The proposed system captures live video streams using a webcam and processes alternate frames to balance computational efficiency with detection continuity. Each frame is analyzed to identify predefined obstacle classes such as pedestrians, vehicles, and bicycles using confidence-based filtering to minimize false detections. To ensure smooth real-time performance, the system incorporates multithreaded processing and asynchronous communication through a Flask-based web framework, enabling stable video streaming and interactive monitoring through a browser interface. Detected objects are annotated with bounding boxes and class labels while detection statistics are dynamically updated on the interface. When safety-critical objects are detected, synchronized audio-visual alerts are automatically triggered to enhance user awareness and response time. Experimental evaluation on a CPU-based platform achieved a detection speed of 14-18 frames per second, 91.3% detection accuracy, and alert latency between 180?240 ms. The results demonstrate that the proposed framework provides reliable real-time performance with efficient resource utilization, making it suitable for practical surveillance and safety monitoring applications.