ADHIYAMAAN COLLEGE OF ENGINEERING
(Autonomous), Hosur
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING
(Accredited by NBA)
MAIN PROJECT
AI-Powered Accident Detection System and Emergency Alerts
ABSTRACT
BATCH: 2024-2025
CLASS: IV – A
The AI-Powered Video Surveillance for Real-Time Accident Detection and SMS Alerts
system leverages deep learning and computer vision to automatically detect vehicular
accidents from video footage and trigger emergency SMS notifications instantly. The core of
this system is a Faster R-CNN model trained on a custom dataset in COCO format, capable
of detecting accidents with high accuracy from live or recorded video streams. Built using
PyTorch and Torchvision, the model applies region-based convolutional feature extraction
and bounding box prediction to locate and classify accident scenarios within video frames.
The system continuously processes video input, overlays detection boxes and confidence
scores on detected events, and sends emergency alerts via Twilio SMS API when accidents
are confidently identified. With retry logic built into the alerting mechanism, the system
ensures reliable communication even in case of temporary failures. Designed for deployment
in intelligent traffic monitoring environments, this solution enhances road safety by enabling
immediate incident detection and response. By combining advanced AI surveillance with
automated alerting, the project aims to reduce response times, minimize accident impact, and
support real-time decision-making in smart transportation systems.
TEAM MEMBERS :
1. GAUTHAM SAIKUMAR K [6176AC21UCS031]
2. HARIKRISHNA P N [6176AC21UCS042]
3. ILANTHENRAL V [6176AC21UCS046]
