Evaluation of synthetic data impact on fire segmentation models performance

Abstract Timely fire detection in industrial environments is crucial to safeguarding people and property. Deep neural networks have shown effectiveness in fire detection over traditional methods. However, they require high-quality datasets, which are costly and time-intensive to gather. To overcome...

Full description

Saved in:
Bibliographic Details
Main Authors: Matej Arlovic, Franko Hrzic, Mitesh Patel, Tomasz Bednarz, Josip Balen
Format: Article
Language:English
Published: Nature Portfolio 2025-05-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-01571-5
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849744689535647744
author Matej Arlovic
Franko Hrzic
Mitesh Patel
Tomasz Bednarz
Josip Balen
author_facet Matej Arlovic
Franko Hrzic
Mitesh Patel
Tomasz Bednarz
Josip Balen
author_sort Matej Arlovic
collection DOAJ
description Abstract Timely fire detection in industrial environments is crucial to safeguarding people and property. Deep neural networks have shown effectiveness in fire detection over traditional methods. However, they require high-quality datasets, which are costly and time-intensive to gather. To overcome this issue, we created the SYN-FIRE dataset, consisting of 2000 labeled images of simulated indoor industrial fires using NVIDIA Omniverse. By using U-Net++ as the baseline, this study explores the impact of the new SYN-FIRE dataset on models’ performance when combined with four publicly available datasets. Two ablation studies were conducted: one replacing portions of real data from publicly available datasets with synthetic data and the other adding various amounts of synthetic data. With over 200 models trained across three resolutions, the results indicate that incorporating additional synthetic data improved DiceScore by $$2.06\%$$ to $$16.09\%$$ (FireBot and BowFire datasets, respectively) while substituting real data with synthetic data generally enhanced performance but with exceptions. Furthermore, tests on challenging real-life fire images confirmed that synthetic data boosts model generalization, supported by GRAD-CAM saliency maps. Finally, we provide key takeaways that point out the main findings of our research. The SYN-FIRE dataset is publicly available to encourage further research in fire detection and prevention.
format Article
id doaj-art-7763aeb2f83547adb2bf3870141e6ad3
institution DOAJ
issn 2045-2322
language English
publishDate 2025-05-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-7763aeb2f83547adb2bf3870141e6ad32025-08-20T03:10:13ZengNature PortfolioScientific Reports2045-23222025-05-0115111410.1038/s41598-025-01571-5Evaluation of synthetic data impact on fire segmentation models performanceMatej Arlovic0Franko Hrzic1Mitesh Patel2Tomasz Bednarz3Josip Balen4University of J.J. Strossmayer Osijek, Faculty of Electrical Engineering, Computer Science and Information TechnologyDepartment of Orthopaedic Surgery and Sports Medicine, Boston Children’s Hospital, Harvard Medical SchoolNVIDIA CorporationNVIDIA CorporationUniversity of J.J. Strossmayer Osijek, Faculty of Electrical Engineering, Computer Science and Information TechnologyAbstract Timely fire detection in industrial environments is crucial to safeguarding people and property. Deep neural networks have shown effectiveness in fire detection over traditional methods. However, they require high-quality datasets, which are costly and time-intensive to gather. To overcome this issue, we created the SYN-FIRE dataset, consisting of 2000 labeled images of simulated indoor industrial fires using NVIDIA Omniverse. By using U-Net++ as the baseline, this study explores the impact of the new SYN-FIRE dataset on models’ performance when combined with four publicly available datasets. Two ablation studies were conducted: one replacing portions of real data from publicly available datasets with synthetic data and the other adding various amounts of synthetic data. With over 200 models trained across three resolutions, the results indicate that incorporating additional synthetic data improved DiceScore by $$2.06\%$$ to $$16.09\%$$ (FireBot and BowFire datasets, respectively) while substituting real data with synthetic data generally enhanced performance but with exceptions. Furthermore, tests on challenging real-life fire images confirmed that synthetic data boosts model generalization, supported by GRAD-CAM saliency maps. Finally, we provide key takeaways that point out the main findings of our research. The SYN-FIRE dataset is publicly available to encourage further research in fire detection and prevention.https://doi.org/10.1038/s41598-025-01571-5Deep learningFire datasetFire detectionIndustrial fireSYN-FIRESynthetic data
spellingShingle Matej Arlovic
Franko Hrzic
Mitesh Patel
Tomasz Bednarz
Josip Balen
Evaluation of synthetic data impact on fire segmentation models performance
Scientific Reports
Deep learning
Fire dataset
Fire detection
Industrial fire
SYN-FIRE
Synthetic data
title Evaluation of synthetic data impact on fire segmentation models performance
title_full Evaluation of synthetic data impact on fire segmentation models performance
title_fullStr Evaluation of synthetic data impact on fire segmentation models performance
title_full_unstemmed Evaluation of synthetic data impact on fire segmentation models performance
title_short Evaluation of synthetic data impact on fire segmentation models performance
title_sort evaluation of synthetic data impact on fire segmentation models performance
topic Deep learning
Fire dataset
Fire detection
Industrial fire
SYN-FIRE
Synthetic data
url https://doi.org/10.1038/s41598-025-01571-5
work_keys_str_mv AT matejarlovic evaluationofsyntheticdataimpactonfiresegmentationmodelsperformance
AT frankohrzic evaluationofsyntheticdataimpactonfiresegmentationmodelsperformance
AT miteshpatel evaluationofsyntheticdataimpactonfiresegmentationmodelsperformance
AT tomaszbednarz evaluationofsyntheticdataimpactonfiresegmentationmodelsperformance
AT josipbalen evaluationofsyntheticdataimpactonfiresegmentationmodelsperformance