Real-Time Crash & Fire Detection from CCTV Streams

YOLOv8 model trained on a custom accident/fire/smoke dataset. Live MJPEG streaming, instant WebSocket alerts with snapshot evidence, and a MongoDB alert log.

View Source on GitHub Run it yourself

✨ Features

🔥 3-class detection

Accident · Fire · Smoke — custom YOLOv8 model (best.pt, 22 MB) trained on the 30_SCFD dataset.

📡 Live streaming

Annotated frames served as an MJPEG feed (/video_feed) — watch detections as they happen.

🚨 Instant alerts

flask-socketio pushes alert cards with base64 snapshot + timestamp. Throttled per-class so one incident doesn't flood the DB.

🔐 Auth + dashboard

Signup/signin with hashed passwords, protected dashboard & alert history pages.

🗄️ Alert log

MongoDB stores every alert (type, image, timestamp); history API returns the latest 100.

🎥 Any video source

Video file, webcam index, or RTSP camera URL — set VIDEO_SOURCE and go.

🏗️ Architecture

CCTV / video file ──► YOLOv8 (custom weights)
                          │
                          ▼ frame inference + annotation
                Flask + flask-socketio ────► browser (MJPEG stream + alert toasts)
                          │
                          ▼
                     MongoDB (alerts collection)

Python 3.10Flaskflask-socketioYOLOv8 / UltralyticsOpenCVMongoDBeventlet

▶️ Quick start

git clone https://github.com/alokscfd9004/crash-fire-detection
cd crash-fire-detection/Final-Year-Project/final_project/Code
pip install -r requirements.txt
cp .env.example .env      # set SECRET_KEY + MONGO_URI
python app.py             # → http://localhost:5500

📄 Full methodology + results are in the project report and published paper (IJSREM) bundled in the repository.

📬 Contact

Built by Alok Kumar & team · portfolio: my-portfolio-alok.vercel.app