🔥 Fire Detection System using MobileNetV2 + OpenCV

📌 Overview

This project is a real-time fire detection system built using MobileNetV2 (TensorFlow/Keras) for deep learning and OpenCV for computer vision.
It can detect fire in images, videos, or live webcam feeds with high accuracy.

🚀 Features

• ✅ Trainable model (binary classification: Fire / No Fire)
• ✅ Preprocessing: image resizing, normalization, augmentation
• ✅ Model optimization with EarlyStopping & ModelCheckpoint
• ✅ Real-time detection with OpenCV (webcam/video feed)
• ✅ Saved model weights for reusability (.h5 files)

🛠️ Tech Stack

• Python 3.10+
• TensorFlow / Keras
• OpenCV
• NumPy, Matplotlib, scikit-learn

📂 Project Structure

project/
│── dataset/ # Training dataset (Fire / NoFire folders)
│ ├── Fire/
│ └── NoFire/
│── fire_detection_model.py # Script for training MobileNetV2 model
│── fire_detection_app.py # Real-time detection using trained model
│── best_fire_model.h5 # Saved trained model (example)
│── fire_detection_model_mobilenet.h5 # Another trained model version
│── requirements.txt # Dependencies
│── .gitignore
│── LICENSE
│── README.md

---

▶️ How to Run

1️⃣ Clone the Repository

git clone https://github.com/your-username/fire-detection-mobilenet.git
cd fire-detection-mobilenet

2️⃣ Setup Virtual Environment

python -m venv venv
venv\Scripts\activate     # On Windows

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Dataset Setup

dataset/
├── Fire/
└── NoFire/

5️⃣ Train the Model

python fire_detection_model.py

This will train MobileNetV2 and save the best model as .h5.

6️⃣ Run Real-Time Fire Detection

python fire_detection_app.py

Opens your webcam/video feed

Detects Fire / No Fire in real time

📊 Results

Model Accuracy: 92%+ on custom dataset

Fast inference with MobileNetV2

Alerts when fire is detected in video stream

👤 Author

Nirjana Das

GitHub:nirjanadas