FireXnet: an explainable AI-based tailored deep learning model for wildfire detection on resource-constrained devices


Ahmad, Khubab and Khan, Muhammad Shahbaz and Ahmed, Fawad and Driss, Maha and Boulila, Wadii and Alazeb, Abdulwahab and Alsulami, Mohammad and Alshehri, Mohammed S. and Ghadi, Yazeed Yasin and Ahmad, Jawad (2023) FireXnet: an explainable AI-based tailored deep learning model for wildfire detection on resource-constrained devices. Fire Ecology, 19 (1). ISSN 1933-9747

[img] Text
39.pdf - Published Version
Restricted to Repository staff only

Download (4MB)


Forests cover nearly one-third of the Earth’s land and are some of our most biodiverse ecosystems. Due to climate change, these essential habitats are endangered by increasing wildfres. Wildfres are not just a risk to the environment, but they also pose public health risks. Given these issues, there is an indispensable need for ef�cient and early detection methods. Conventional detection approaches fall short due to spatial limitations and man�ual feature engineering, which calls for the exploration and development of data-driven deep learning solutions. This paper, in this regard, proposes ’FireXnet’, a tailored deep learning model designed for improved efciency and accu�racy in wildfre detection. FireXnet is tailored to have a lightweight architecture that exhibits high accuracy with sig�nifcantly less training and testing time. It contains considerably reduced trainable and non-trainable parameters, which makes it suitable for resource-constrained devices. To make the FireXnet model visually explainable and trust�able, a powerful explainable artifcial intelligence (AI) tool, SHAP (SHapley Additive exPlanations) has been incorpo�rated. It interprets FireXnet’s decisions by computing the contribution of each feature to the prediction. Furthermore, the performance of FireXnet is compared against fve pre-trained models — VGG16, InceptionResNetV2, InceptionV3, DenseNet201, and MobileNetV2 — to benchmark its efciency. For a fair comparison, transfer learning and fne-tun�ing have been applied to the aforementioned models to retrain the models on our dataset. Results The test accuracy of the proposed FireXnet model is 98.42%, which is greater than all other models used for comparison. Furthermore, results of reliability parameters confrm the model’s reliability, i.e., a confdence interval of [0.97, 1.00] validates the certainty of the proposed model’s estimates and a Cohen’s kappa coefcient of 0.98 proves that decisions of FireXnet are in considerable accordance with the given data. Conclusion The integration of the robust feature extraction of FireXnet with the transparency of explainable AI using SHAP enhances the model’s interpretability and allows for the identifcation of key characteristics triggering wildfre detections. Extensive experimentation reveals that in addition to being accurate, FireXnet has reduced computa�tional complexity due to considerably fewer training and non-training parameters and has signifcantly fewer training and testing times.

Item Type: Article
Uncontrolled Keywords: Wildfre, Fire detection, CNN, Transfer learning, Lightweight architecture
Subjects: T Technology > TH Building construction > TH9025-9745 Protection of buildings Including protection from dampness, fire, burglary
Divisions: Faculty of Engineering and Technology (FET)
Depositing User: Ms Nurul Iqtiani Ahmad
Date Deposited: 31 Oct 2023 08:05
Last Modified: 31 Oct 2023 08:05


Downloads per month over past year

View ItemEdit (login required)