Fire Detection Using Machine Learning

Abstract

Fire accidents cause significant damage to life, property, and the environment, making early detection essential for safety and prevention. This project focuses on developing a fire-detection system using Machine-Learning techniques to identify fire and smoke in real time. The system utilises image processing and deep-learning models, particularly Convolutional Neural Networks (CNN), to analyse visual features such as colour, texture, shape, and motion patterns associated with fire. A dataset consisting of fire and non-fire images is collected and preprocessed to train the model effectively, and the trained model is capable of accurately classifying images and detecting fire in video streams captured through surveillance cameras. When fire is detected, the system can trigger alerts such as alarms or notifications to ensure a quick response. Compared with traditional sensor-based methods, this ML-based approach offers improved accuracy, faster response time, and reduced false alarms, and enables continuous monitoring without human intervention. The prototype is implemented in Python using TensorFlow/Keras for the CNN, OpenCV for image and video processing, and a Flask or Django web interface for uploading images or video and viewing detection results. The system was validated through nineteen functional, validation, and performance test cases covering application launch, image and video upload, invalid input, model execution, fire/no-fire classification, alert generation, and real-time response, all of which passed. The proposed system can be applied in residential, industrial, and forest environments, providing a reliable and cost-effective solution for early fire detection and disaster management.