FLAME

该数据集名为FLAME，由在美国亚利桑那州北部可控堆烧期间无人机拍摄的图像组成。它包含了带有火焰的图像（火警图像）和其他不包含任何火焰的图像（非火警图像），以及由无人机拍摄的视频中的连续帧。该数据集展现了显著相似性的多种图像，使得能够评估FL数据集预期的非独立同分布行为。本地数据集被划分为80%用于训练，20%用于测试。相关任务是联邦学习参与者选择。

This dataset, named FLAME, consists of images captured by drones during controlled pile burns in northern Arizona, USA. It includes two categories of images: those containing flames (fire images) and those without any flames (non-fire images), as well as consecutive frames extracted from drone-captured videos. This dataset features multiple groups of images with notable similarity, enabling the assessment of the expected non-independent and identically distributed (non-IID) behavior of federated learning (FL) datasets. The local dataset is split into 80% for training and 20% for testing. The relevant task is federated learning participant selection.