Fire’s Latent Activity Monitoring and Evaluation through Thermography (FLAME-T)

This repository presents FLAME-T, an annotated dataset of a prescribed burn conducted in Gran Canaria. This dataset is structured to aid in thermal imaging analysis for fire monitoring and evaluation. Dataset Composition: The FLAME-T dataset consists of approximately 2,640 thermal images, each captured with high precision using a terrestrial thermal camera system. The dataset was annotated using the You Only Look Once (YOLO) format, allowing for easy integration with object detection algorithms and machine learning models focused on fire behavior analysis and other related applications. Technical Specifications: File Format: JPEG, TXT Resolution: Ranges from 160x120 to 336x256 Cameras Used: FLIR Tau 2, FLIR A35, and FLIR Lepton Image Capture Protocol: The thermal images were captured from multiple geographic locations and varying perspectives of the same fire, providing a comprehensive view of the fire’s progression and intensity across different vantage points. Images are grouped in batches of 20 frames, each captured at 1-second intervals from a single perspective, enabling fine-grained temporal analysis of the fire’s latent activity. Geographical Information: Fire location: 27.990619, -15.518103, 737 m Point A: 27.997256, -15.514588, 768 m Point B: 27.997955, -15.510110, 789 m Point C: 27.995225, -15.505335, 769 m Point D: 27.992816, -15.521879, 861 m Point E: 28.000828, -15.517579, 971 m Point F: 27.999762, -15.515277, 951 m These coordinates allow users to accurately understand the location and elevation context of each image batch, facilitating spatial analysis of fire activity across different terrains. Metadata Folder: A folder named metadata includes supplemental meteorological information (e.g., temperature, humidity, wind speed) gathered using the Open-Meteo API, providing environmental context for each image batch. Attached to this repository is the dataset created and used during the work. If you wish to use the code sample to replicate our work, please use the link found in the 'Related links' section, which leads directly to the official GitHub of the paper.本仓库收录了FLAME-T数据集，该数据集是在大加那利岛开展的计划烧除（prescribed burn）的标注数据集，其构建目的是为火灾监测与评估领域的热成像分析提供支持。 数据集构成： FLAME-T数据集包含约2640张热成像图像，所有图像均通过地面热成像相机系统高精度采集。该数据集采用你只看一次（YOLO）格式进行标注，可轻松集成至面向火灾行为分析及其他相关应用的目标检测算法与机器学习模型中。 技术规格： 文件格式：JPEG、TXT 分辨率：160×120至336×256不等 所用相机：FLIR Tau 2、FLIR A35及FLIR Lepton 图像采集协议： 本数据集的热成像图像采集自同一场火灾的多个地理位置与不同观测视角，可全面呈现火灾在不同观测点下的蔓延过程与强度变化。图像以20帧为一组进行批量存储，每组内的图像均以1秒为间隔从同一视角采集，支持对火灾潜在活动进行细粒度的时序分析。 地理信息： 火灾中心点坐标：27.990619, -15.518103，海拔737米 点位A：27.997256, -15.514588，海拔768米 点位B：27.997955, -15.510110，海拔789米 点位C：27.995225, -15.505335，海拔769米 点位D：27.992816, -15.521879，海拔861米 点位E：28.000828, -15.517579，海拔971米 点位F：27.999762, -15.515277，海拔951米 上述坐标可帮助使用者精准掌握各图像组的位置与海拔环境，便于开展不同地形下火灾活动的空间分析。 元数据文件夹： 名为metadata的文件夹包含通过Open-Meteo API采集的补充气象数据（如气温、湿度、风速），可为各图像组提供环境背景信息。 本仓库附带本研究中创建并使用的数据集。若需使用代码示例复现本研究工作，请访问“相关链接”板块中的链接，该链接直接指向本论文的官方GitHub仓库。