| Title |
Optimizing fire detection in remote sensing imagery for edge devices: a quantization-enhanced hybrid deep learning model |
| Authors |
Bukhari, Syed Muhammad Salman ; Dahmani, Nadia ; Gyawali, Sujan ; Zafar, Muhammad Hamza ; Sanfilippo, Filippo ; Raja, Kiran |
| DOI |
10.1016/j.displa.2025.103070 |
| Full Text |
|
| Is Part of |
Displays.. Amsterdam : Elsevier. 2025, vol. 89, art. no. 103070, p. 1-12.. ISSN 0141-9382. eISSN 1872-7387 |
| Keywords [eng] |
Bushfire detection ; Inception-resNet ; Quantization ; Smart city applications ; Transformer models ; Unmanned aerial vehicles (UAV) |
| Abstract [eng] |
Wildfires are increasing in frequency and severity, presenting critical challenges for timely detection and response, particularly in remote or resource-limited environments. This study introduces the Inception-ResNet Transformer with Quantization (IRTQ), a novel hybrid deep learning (DL) framework that integrates multi-scale feature extraction with global attention and advanced quantization. The proposed model is specifically optimized for edge deployment on platforms such as unmanned aerial vehicles (UAVs), offering a unique combination of high accuracy, low latency, and compact memory footprint. The IRTQ model achieves 98.9% accuracy across diverse datasets and shows strong generalization through cross-dataset validation. Quantization significantly reduces the parameter count to 0.09M and memory usage to 0.13 MB, enabling real-time inference in 3 ms. Interpretability is further enhanced through Grad-CAM visualizations, supporting transparent decision-making. While achieving state-of-the-art performance, the model encounters challenges in visually ambiguous fire-like regions. To address these, future work will explore multi-modal inputs and extend the model towards multi-class classification. IRTQ represents a technically grounded, interpretable, and deployable solution for AI-driven wildfire detection and disaster response. |
| Published |
Amsterdam : Elsevier |
| Type |
Journal article |
| Language |
English |
| Publication date |
2025 |
| CC license |
|
