ESA-PhiLab-Edge/LPL-Burned-Area-Seg

PhiSat-2火烧区域数据集基于Sentinel-2影像，提供了4.75米空间分辨率的模拟火烧区域数据，可用于全球不同生态系统的火灾后景观详细分析。数据集包含115次野火事件，覆盖6个大陆，包括 boreal、温带和热带三种火灾类型。每个影像都被标记为四个类别：背景、火烧区域、云层和水体。这些类别的分布有助于模型处理现实世界中的问题，如云层覆盖和外观相似的特性，提高模型的性能和泛化能力。数据集在测试、验证和训练集中的类别分布均衡，非常适合用于野火映射、监控和生态学研究。

The PhiSat-2 Burned Area Dataset is based on Sentinel-2 imagery and provides simulated burned area data at a 4.75-meter spatial resolution, allowing for detailed post-fire landscape analysis in various ecosystems around the world. The dataset includes 115 wildfire events, covering 6 continents and including boreal, temperate, and tropical fire types. Each image is labeled with four classes: Background, Burned Area, Clouds, and Waterbodies. The distribution of these classes helps models handle real-world issues like cloud cover and similar-looking features, improving model performance and generalization. The dataset has a balanced class distribution across the test, validation, and training sets, making it well-suited for wildfire mapping, monitoring, and ecological research.