CS619 Final Year Project — Spring 2026

FireGuard AI 🔥

An Automated Fire Detection System powered by a custom Convolutional Neural Network (TensorFlow/Keras) and the Django web framework. Classifies images as Fire or Non-Fire through live webcam feeds, IP cameras, and uploaded images — with a server-side sound alarm on detection.

Shoaib Ejaz

BS Computer Science — Final Year Student

Virtual University of Pakistan
Course: CS619 — Final Year Project
Supervisor: Madiha Faqir Hussain
Student ID: (add your VU student ID here)
Semester: Spring 2026

🎯

Project Goal

Detect fire and smoke automatically using AI — faster and more reliably than traditional smoke detectors.

🧠

AI Model

Custom CNN built with TensorFlow/Keras. Trained on 10,003 images (Fire / Non-Fire). Test accuracy: 92.60%.

⚡

Runs Anywhere

Classifies 128×128 frames on CPU — no GPU required. Live feed uses consecutive-frame smoothing to cut down false alarms.

Technology Stack

🐍 Django Backend framework

🤖 TensorFlow / Keras CNN classifier

📷 OpenCV Video & image processing

🎨 Bootstrap 5 Frontend UI

🗄️ SQLite Database

📊 Chart.js Dashboard charts

🔊 pygame Server-side sound alarm

🌐 Nginx + Gunicorn Production deployment

System Features

✓ Fire detection on user-uploaded images (PNG/JPG/JPEG), with an annotated result banner and confidence score
✓ Live webcam feed detection with consecutive-frame smoothing to reduce false alarms
✓ IP camera support — add, list, stream, and delete RTSP/HTTP camera sources
✓ Server-side sound alarm (pygame) triggered automatically on fire detection
✓ Dashboard with live stats, a 7-day detection trend chart, and a Fire vs Safe breakdown
✓ Per-user detection history log
✓ User authentication — registration, login, and logout
✓ Django admin panel for managing detections and cameras