@article{ART003383525},
author={Miseo Kim and Eun-Jin Kim and Young-Seob Jeong},
title={An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={9},
pages={151-161}
TY - JOUR
AU - Miseo Kim
AU - Eun-Jin Kim
AU - Young-Seob Jeong
TI - An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 9
PB - The Korean Society Of Computer And Information
SP - 151
EP - 161
SN - 1598-849X
AB - This study proposes a DINOv2-Swin ensemble model based on Transformers to improve outdoor fire detection accuracy and address the persistent issue of Unwanted Fire Alarms (UFAs). Outdoor environments pose challenges for precise detection due to vast areas and complex backgrounds, leading to high false alarm rates. To overcome this, we combined the strengths of DINOv2 for global visual feature extraction and the Swin Transformer for localized pattern recognition. To evaluate practical applicability, 1-FPR — i.e., specificity (true negative rate), computed as one minus the false positive rate — was adopted as a key metric alongside accuracy to assess the suppression of false alarms.
Experimental results using the FLAME (Fire Luminosity Analysis in Multispectral Images) dataset demonstrated that the DINOv2-Swin model achieved an accuracy of 81.21% and a 1-FPR of 94.31%, representing a 4.98%p improvement in accuracy over the previous Xception-based approach.
Furthermore, additional evaluation on the external D-Fire dataset provided evidence of the model's cross-dataset generalization potential. Overall, this research suggests that an image-based Transformer ensemble can help reduce UFAs, with potential applicability to firefighting response efficiency and future intelligent fire surveillance systems.
KW - Outdoor Fire Detection;Transformer Ensemble;DINOv2;Swin Transformer;Unwanted Fire Alarms
DO -
UR -
ER -
Miseo Kim, Eun-Jin Kim and Young-Seob Jeong. (2026). An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms. Journal of The Korea Society of Computer and Information, 31(9), 151-161.
Miseo Kim, Eun-Jin Kim and Young-Seob Jeong. 2026, "An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms", Journal of The Korea Society of Computer and Information, vol.31, no.9 pp.151-161.
Miseo Kim, Eun-Jin Kim, Young-Seob Jeong "An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms" Journal of The Korea Society of Computer and Information 31.9 pp.151-161 (2026) : 151.
Miseo Kim, Eun-Jin Kim, Young-Seob Jeong. An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms. 2026; 31(9), 151-161.
Miseo Kim, Eun-Jin Kim and Young-Seob Jeong. "An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms" Journal of The Korea Society of Computer and Information 31, no.9 (2026) : 151-161.
Miseo Kim; Eun-Jin Kim; Young-Seob Jeong. An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms. Journal of The Korea Society of Computer and Information, 31(9), 151-161.
Miseo Kim; Eun-Jin Kim; Young-Seob Jeong. An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms. Journal of The Korea Society of Computer and Information. 2026; 31(9) 151-161.
Miseo Kim, Eun-Jin Kim, Young-Seob Jeong. An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms. 2026; 31(9), 151-161.
Miseo Kim, Eun-Jin Kim and Young-Seob Jeong. "An Ensemble Approach for Outdoor Fire Detection Using DINOv2 and Swin Transformer to Minimize Unwanted Fire Alarms" Journal of The Korea Society of Computer and Information 31, no.9 (2026) : 151-161.
