Convolutional Neural Networks: Comparative Fire Detection Using Deep Learning Algorithm

Abstract

Recent studies have showcased significant accomplishments in leveraging computer vision and deep learning techniques, significantly contributing to the realm of fire detection. Among these techniques, computer vision and AI-driven approaches, including static and dynamic texture analysis. A major safety problem for both buildings and public areas is fire detection. The reliability and coverage of traditional approaches depending on sensor networks are constrained. Convolutional Neural Networks, also known as CNNs, are a subset of artificial neural networks used in deep learning and are frequently employed for object and picture recognition and categorization. Thus, Deep Learning uses a CNN to identify items in a picture. There is a significant scarcity of publicly available datasets for fire detection in images, the factors that influenced our choices to use the Flame Vision dataset from Kaggle for this research work development are sizes, annotations, and balanced classes. The Flame Vision dataset consists of 8600 complete high resolution photographs. A purposeful and intentional divide has been made within this collection: 5000 photographs are devoted to exhibiting diverse fire scenarios, and the remaining 3600 images show scenes without any fires. This paper suggests that CNN and transfer learning models using VGG-16 offer the most promising results for the task of wildfire image classification. These models not only yield higher accuracy but also demonstrate a more balanced trade-off between precision and recall compared to traditional ANN.

Keywords

Convolutional Neural Networks Fire Detection Deep Learning Comparative Algorithm

Suggested Citation

Meju Moses, Ekoko Ujerekre, Chiadika Dominic Mario. (2026). Convolutional Neural Networks: Comparative Fire Detection Using Deep Learning Algorithm. Academy Journal of Information and Communication Technology, 1(1).

Download Full Article PDF