MultiFire20K Dataset

Overview

MultiFire20K is a UAV-based dataset for fire monitoring, designed for image classification, segmentation, and multi-task learning. It contains 20,500 images extracted from 67 UAV videos captured across diverse urban and rural environments worldwide.

Statistics

Split	Images
Train	14,350
Validation	3,075
Test	3,075
Total	20,500

Classes

• Fire
• Normal

Environment Types

• Urban (Build-Up)
• Rural (Natural Landscape)

Source

Original dataset:

https://zenodo.org/records/17047113

Citation

@article{shianios2025multifire20k,
  title={MultiFire20K: A semi-supervised enhanced large-scale UAV-based benchmark for advancing multi-task learning in fire monitoring},
  author={Shianios, Demetris and Kolios, Panayiotis and Kyrkou, Christos},
  journal={Computer Vision and Image Understanding},
  volume={254},
  pages={104318},
  year={2025},
  publisher={Elsevier}
}