kostaspic/amfitrite-open-waters-hab-sentinel2
收藏资源简介:
Amfitrite-Open-Waters-HAB-Sentinel2是一个专门用于检测和分类开放水域中有害藻华(HABs)的数据集。该数据集包含2,982个来自Sentinel-2卫星图像的多光谱图像块,分为1,445个HAB块、1,317个非HAB块、110个纯云干扰块和110个纯陆地干扰块。注释方法结合了来自9个不同全球数据集的现场细胞计数测量、空间网格采样以减轻坐标不准确性、CyFi工具的预测以及针对4个不同深度学习模型的交叉验证。该数据集旨在训练机器学习或深度学习模型(如CNNs),用于大规模环境监测和海洋异常检测。
Amfitrite-Open-Waters-HAB-Sentinel2 is a specialized dataset designed for the detection and classification of Harmful Algal Blooms (HABs) in open waters. The dataset consists of 2,982 multispectral image tiles derived from Sentinel-2 satellite imagery, featuring 1,445 HAB tiles, 1,317 non-HAB tiles, 110 pure cloud distractors, and 110 pure land distractors. The annotation methodology integrates in-situ cell count measurements from 9 different global datasets, spatial grid sampling to mitigate coordinate inaccuracies, predictions from the CyFi tool, and cross-validation against 4 distinct Deep Learning models. It is designed to train Machine or Deep Learning models (like CNNs) for large-scale environmental monitoring and ocean anomaly detection.




