FlaDE Dataset

Overview

FlaDE (Event‑Camera‑Based Flame Detection Dataset) is a dataset specifically designed for event‑based flame detection. Traditional RGB cameras perform poorly when dealing with static backgrounds, over‑exposure, and redundant data, while event cameras offer a biologically inspired alternative that effectively addresses these challenges, making them especially suitable for flame‑detection tasks.

Contents

• Event data and annotation files: Interfaces are provided for reading event data and annotation files.
• Detection examples: Example code for flame detection.
• Evaluation tools: Tools for evaluating detection results.
• Visualization utilities: Support for visualizing the dataset.

Usage

Install Dependencies

1. Install required packages:

   bash setup.sh
   

2. Create a Conda environment:

   conda create -n cocoa python=3.8
   

3. Install necessary packages:

   conda activate cocoa
   pip install -r requirements.txt
   pip install external/dv-toolkit/.
   pip install cocoa_flade/.
   

Read Data

Use the following Python script to read FlaDE data:

import cocoa_flade as cocoa
dataset = cocoa.FlaDE(<file_path>)
cats = dataset.get_cats(key=name, query=None)
tags = dataset.get_tags(key=partition, query=[train, val])

Run Example

Execute the real‑time flame detection example based on event cameras:

conda activate cocoa
cd ./samples/bec_svm
bash setup.sh
python3 samples/bec_svm/demo.py

Citation

If you use this dataset, please cite the following paper:

@article{ding2024hyper,
  title={Hyper real‑time flame detection: Dynamic insights from event cameras and FlaDE dataset},
  author={Ding, Saizhe and Zhang, Haorui and Zhang, Yuxin and Huang, Xinyan and Song, Weiguo},
  journal={Expert Systems with Applications},
  volume={263},
  pages={125764},
  year={2024},
  publisher={Elsevier},
}