Intelligent Fire and Smoke Detection Using Deep Learning and MobileNet Architecture

Vinuth N, Dr. Durgha Devi P

Abstract: This research paper presents a real-time intelligent fire and smoke detection system designed using deep learning and the MobileNet architecture. The system utilizes computer vision and convolutional neural networks (CNN) to accurately detect fire and smoke from images, recorded videos, and live webcam feeds. Unlike traditional hardware-based detectors, this approach offers early visual detection, scalability, and flexibility for diverse environments. Trained on a dataset of 3,825 images, the system achieved 97% training accuracy and 94% validation accuracy. Implemented using Python, TensorFlow, OpenCV, and Flask, the model provides reliable alerts in real time. The proposed solution demonstrates strong potential for smart city, industrial, and environmental applications.

Keywords: Fire Detection, Smoke Detection, Deep Learning, MobileNet, CNN, Flask, Computer Vision

How to Cite?: Vinuth N, Dr. Durgha Devi P, "Intelligent Fire and Smoke Detection Using Deep Learning and MobileNet Architecture", Volume 14 Issue 12, December 2025, International Journal of Science and Research (IJSR), Pages: 1810-1811, https://www.ijsr.net/getabstract.php?paperid=SR251222150552, DOI: https://dx.doi.org/10.21275/SR251222150552