International Journal of Science, Engineering and TechnologyAuthors: Mr. M. V. Rajesh, Adapa Amrutha Varshini, Sabbarapu Kumar Ganesh, Mutyala Sowmya, Koppisetti N V Prasanna Sri Sandeep, Kanupudi Mahathi Sree
Abstract: Early fire detection plays a vital role in preventing major disasters, reducing property loss, and ensuring public safety. Conventional fire detection systems primarily depend on physical sensors such as smoke, heat, and gas detectors. While these methods are commonly used, they often face challenges such as delayed response, high false alarm rates, and reduced effectiveness in complex environments like industrial areas and densely populated urban regions. With the rapid growth of computer vision and deep learning technologies, image-based intelligent fire detection systems have emerged as a promising solution for early detection. This study presents a deep learning-based fire detection and early warning system that utilizes Convolutional Neural Networks (CNN) to automatically detect fire from images captured through surveillance cameras. The proposed model extracts visual features from images and classifies them into two categories: fire and non-fire. A well-structured dataset consisting of fire and non-fire images is used for training and validation of the model. To enhance generalization and minimize overfitting, data augmentation techniques such as rotation, scaling, and horizontal flipping are applied. Additionally, optimization methods including Early Stopping and ReduceLROnPlateau are incorporated to improve training efficiency and model stability. The experimental findings indicate that the CNN-based model performs significantly better than traditional machine learning approaches such as Logistic Regression, K-Nearest Neighbor (KNN), and AdaBoost. The proposed system achieves high classification accuracy along with strong recall and AUC values. Moreover, an automated alert mechanism is integrated into the system, which triggers an alarm upon detecting fire, enabling quick emergency response. Overall, the proposed approach offers a cost-effective, reliable, and scalable solution for fire detection, suitable for deployment in surveillance systems across buildings, industrial sectors, and smart city environments. The results demonstrate that deep learning-based visual fire detection systems can greatly improve safety monitoring and disaster prevention.
DOI:
