FireFly A Synthetic Dataset for Ember Detection in Wildfire

TL;DR

FireFly is a synthetic dataset created with Unreal Engine 4 for improved ember detection in wildfires, enabling better model training and real-world performance.

Contribution

The paper introduces a customizable synthetic dataset and a semi-automatic labeling tool for ember detection, addressing data scarcity in wildfire research.

Findings

Up to 8.57% improvement in mAP on real wildfire data.

Generated 19,273 frames for diverse training.

Enhanced model performance with synthetic data.

Abstract

This paper presents "FireFly", a synthetic dataset for ember detection created using Unreal Engine 4 (UE4), designed to overcome the current lack of ember-specific training resources. To create the dataset, we present a tool that allows the automated generation of the synthetic labeled dataset with adjustable parameters, enabling data diversity from various environmental conditions, making the dataset both diverse and customizable based on user requirements. We generated a total of 19,273 frames that have been used to evaluate FireFly on four popular object detection models. Further to minimize human intervention, we leveraged a trained model to create a semi-automatic labeling process for real-life ember frames. Moreover, we demonstrated an up to 8.57% improvement in mean Average Precision (mAP) in real-world wildfire scenarios compared to models trained exclusively on a small real dataset.