Fire and Smoke Detection

This project aims to detect and monitor fire incidents by training a deep learning model to classify fire and smoke images.

Installation

if you are interested in running the project on your device, make sure to install all the dependencies. You can write the command below to install it all:

pip install -r requirements.txt

Data

This project used Fire-Smoke-Dataset which contains 3 classes (i.e., Fire, Smoke, Neutral) with the same proportion of each class (balanced dataset.)

Experiement set-up

Data Preparation

• Data Partitioning: splits the dataset into train/val/test folders in the Data Preparation.ipynb notebook.
• Data Preprocessing: resize, rescale, and augment the dataset in the code.ipynb notebook.

Methodology

The models used in this experiment were:

• CNN model from scratch.
• EfficientNetv2 model.

Results

1. CNN Model

Classification Report:

	Precision	Recall	F1-Score	Overall Accuracy
Fire	0.88	0.83	0.86	0.74
Neutral	0.59	0.87	0.70	-
Smoke	0.88	0.52	0.65	-

Confusion Matrix:

2. EfficicentNetv2

Classification Report:

	Precision	Recall	F1-Score	Overall Accuracy
Fire	0.97	0.92	0.94	0.9633
Neutral	0.99	0.99	0.99	-
Smoke	0.93	0.98	0.96	-

Confusion Matrix:

Deployment

To deploy our model, we used Gradio library to build an interface and host it on HuggingFace.
You can check Fire-Smoke-Detector