HySpecNet-11k
收藏资源简介:
HySpecNet-11k是由柏林工业大学和柏林学习与数据基础研究所创建的大型高光谱数据集,包含11,483个非重叠图像块,每个图像块大小为128×128像素,覆盖224个光谱带,地表采样距离为30米。该数据集通过EnMAP卫星采集,经过辐射、几何和大气校正处理。HySpecNet-11k主要用于学习和评估基于学习的高光谱图像压缩方法,也可用于任何无监督学习任务。数据集的创建解决了现有数据集在训练和评估学习型压缩方法时不足的问题,推动了高光谱图像分析领域的研究。
HySpecNet-11k is a large-scale hyperspectral dataset created by Technische Universität Berlin and the Berlin Institute for Learning and Data Foundations. It comprises 11,483 non-overlapping image patches, each with a size of 128 × 128 pixels, spanning 224 spectral bands and featuring a ground sampling distance of 30 meters. The dataset was acquired via the EnMAP satellite and subjected to radiometric, geometric, and atmospheric corrections. HySpecNet-11k is primarily intended for learning and evaluating learning-based hyperspectral image compression methods, and can also be adapted for any unsupervised learning tasks. The development of this dataset addresses the shortcomings of existing datasets for training and evaluating learning-based compression approaches, thereby advancing research in the field of hyperspectral image analysis.




