FIgLib_ustc_labeled
Overview
FIgLib_ustc_labeled is an object-detection annotation extension of the original Fire Ignition Library (FIgLib) dataset. The original FIgLib dataset was introduced for real-time wildland fire smoke detection using fixed-view camera imagery from Southern California.
Based on the original FIgLib training split, FIgLib_ustc_labeled provides newly annotated wildfire smoke bounding boxes in PASCAL VOC XML format. These annotations convert the original classification-oriented training data into a detection-ready dataset, enabling object detection models to learn smoke localization rather than only image-level smoke presence.
Key Features
- Built on the training split of the original FIgLib dataset.
- Provides wildfire smoke bounding-box annotations.
- Annotation files are provided in PASCAL VOC XML format.
- Supports object detection training for wildfire smoke localization.
- Extends the original classification-oriented FIgLib training data to a detection-oriented setting.
- The original FIgLib validation and test splits are not modified in this release.
Citation
If you use this dataset, please cite the original FIgLib paper:
@Article{rs14041007,
AUTHOR = {Dewangan, Anshuman and Pande, Yash and Braun, Hans-Werner and Vernon, Frank and Perez, Ismael and Altintas, Ilkay and Cottrell, Garrison W. and Nguyen, Mai H.},
TITLE = {FIgLib & SmokeyNet: Dataset and Deep Learning Model for Real-Time Wildland Fire Smoke Detection},
JOURNAL = {Remote Sensing},
VOLUME = {14},
YEAR = {2022},
NUMBER = {4},
ARTICLE-NUMBER = {1007},
URL = {https://www.mdpi.com/2072-4292/14/4/1007},
ISSN = {2072-4292},
DOI = {10.3390/rs14041007}
}and,
@article{https://doi.org/10.1155/int/1610145,
author = {Wang, Chong and Xu, Chen and Akram, Adeel and Wang, Zhong and Shan, Zhilin and Zhang, Qixing},
title = {Wildfire Smoke Detection System: Model Architecture, Training Mechanism, and Dataset},
journal = {International Journal of Intelligent Systems},
volume = {2025},
number = {1},
pages = {1610145},
year = {2025}
}