Deep Learning Based Wildfire Detection for Peatland Fires Using Transfer Learning

Emadeldeen Hamdan 1, Ahmad Faiz Tharima 2, Mohd Zahirasri Mohd Tohir 3, Dayang Nur Sakinah Musa 4, Erdem Koyuncu 1, Adam J. Watts 5, and Ahmet Enis Cetin 1 1 Department of Electrical and Computer Engineering, University of Illinois Chicago, Chicago, IL, USA 2 Fire and Rescue Department of Malaysia, Malaysia 3 Dept. of Chemical and Environmental Engineering, Faculty of Engineering, Universiti Putra Malaysia, Malaysia. 4 International Tropical Forestry Programme, Faculty of Tropical Forestry, Universiti Malaysia Sabah, Malaysia 5 USDA Forest Service Pacific Wildland Fire Sciences Laboratory, Washington, USA · 2026

Read Deep Learning Based Wildfire Detection for Peatland Fires Using Transfer Learning by Emadeldeen Hamdan 1, Ahmad Faiz Tharima 2, Mohd Zahirasri Mohd Tohir 3, Dayang Nur Sakinah Musa 4, Erdem Koyuncu 1, Adam J. Watts 5, and Ahmet Enis Cetin 1 1 Department of Electrical and Computer Engineering, University of Illinois Chicago, Chicago, IL, USA 2 Fire and Rescue Department of Malaysia, Malaysia 3 Dept. of Chemical and Environmental Engineering, Faculty of Engineering, Universiti Putra Malaysia, Malaysia. 4 International Tropical Forestry Programme, Faculty of Tropical Forestry, Universiti Malaysia Sabah, Malaysia 5 USDA Forest Service Pacific Wildland Fire Sciences Laboratory, Washington, USA (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-02465.txt.

Description

Emadeldeen Hamdan, Ahmad Faiz Tharima, Mohd Zahirasri Mohd Tohir, and colleagues adapt wildfire detection to peatland fires in this 2026 paper. Peat fires smolder with low flames, thick smoke, and burning below the surface. Detectors trained on open forest flames miss those signs easily. Labeled peat fire data is scarce, which blocks training from scratch.

Their approach starts from a general wildfire detector's pretrained weights. Fine-tuning on Malaysian peatland images and videos then adapts it to the new domain. Transfer learning beat training from scratch clearly in their experiments. Gains held under low-contrast smoke, partial blocking, and changing light. The method offers a practical path to early peat fire warning. Real-time monitoring for prevention and environmental protection is the target use.

Catalog record

CREATOREmadeldeen Hamdan 1, Ahmad Faiz Tharima 2, Mohd Zahirasri Mohd Tohir 3, Dayang Nur Sakinah Musa 4, Erdem Koyuncu 1, Adam J. Watts 5, and Ahmet Enis Cetin 1 1 Department of Electrical and Computer Engineering, University of Illinois Chicago, Chicago, IL, USA 2 Fire and Rescue Department of Malaysia, Malaysia 3 Dept. of Chemical and Environmental Engineering, Faculty of Engineering, Universiti Putra Malaysia, Malaysia. 4 International Tropical Forestry Programme, Faculty of Tropical Forestry, Universiti Malaysia Sabah, Malaysia 5 USDA Forest Service Pacific Wildland Fire Sciences Laboratory, Washington, USA
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE16350
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.02465

Browse the full library · About the Tower of Babel