The article reviews benchmark remote sensing datasets essential for evaluating learning models. Specifically, it focuses on hyperspectral image datasets such as Pavia, which includes 42,776 images with urban classifications, and Indian Pines, which features multidimensional hyperspectral images from Indiana landscapes for comprehensive multi-label classification. These datasets are not only pivotal in testing model effectiveness but also serve as established standards for researchers aiming to enhance algorithms in the remote sensing domain. The structured classification systems set by these datasets facilitate improved learning and evaluation methods in the field.
The Pavia dataset, designed for multi-label classification, consists of hyperspectral images comprising 610 × 610 pixels and 103 spectral bands, totaling 42,776 labeled images.
Indian Pines, a hyperspectral image dataset focused on a landscape in Indiana, supports multi-label classification with images of 145 × 145 pixels and 16 semantic labels available.
Collection
[
|
...
]