Deep Learning Based Video Surveillance System

This work presents a computer vision and deep learning based real-time vehicle crash detection system. Evaluated for crash event detection from video feeds are four object detection models: YOLOv8, SSD, Faster R-CNN, and EfficientDet. YOLOv8 is chosen as the main model and ideal for live use since it combines speed and accuracy. The system's connection with Twilio's WhatsApp API sends automatic alerts and media evidence upon a crash detection, so increasing its relevance for traffic control and public safety. Live video input makes constant surveillance feasible; experimental comparisons help to evaluate every model's performance. This work presents an intelligent transportation system with scalable solution. Among next improvements are edge deployment, multi-angle video support, and crash intensity classification. Keywords: Deep Learning, Vehicle Crash Detection, YOLOv8, SSD, Faster R-CNN, EfficientDet, WhatsApp Alert, Twilio API, Smart Surveillance, Real-time Detection, Computer Vision