Karst Depression Dataset

The dataset consists of images and corresponding labels designed to study karst landscapes. Each image in the dataset is a composite patch composed of multiple layers or spectral bands, encapsulating a wealth of geographical and topographical information. These patches have dimensions of 128 pixels by 128 pixels, with each pixel representing a specific area on the ground. The dataset contains a total of 55 spectral bands, organized as follows:

- SRTM (Shuttle Radar Topography Mission) Bands (0-10): These bands are derived from the SRTM mission and provide elevation data, crucial for identifying karst features such as sinkholes and ridges. The SRTM data offers a foundational layer for understanding the surface topography.

- ASTER GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model) Bands (11-21): These bands offer additional elevation data with a focus on thermal emissions and reflection, which helps in differentiating between various landforms and can be particularly useful in karst terrain analysis.

- AW3D30 (ALOS World 3D - 30m) Bands (22-32): Derived from the ALOS (Advanced Land Observing Satellite), these bands provide high-resolution elevation data, enhancing the detail with which karst features can be identified and analyzed.

- GLO-30 (Global Digital Elevation Model - 30m) Bands (33-43): GLO-30 bands offer a global perspective on elevation, sourced from a compilation of high-resolution satellite images, aiding in a comprehensive understanding of karst landscapes on a broader scale.

- NASADEM (NASA Digital Elevation Model) Bands (44-54): These bands are the latest in elevation modeling from NASA, offering refined elevation data that incorporates data from previous missions like SRTM, but with improved accuracy and resolution.

A label with dimensions of 128x128 accompanies each image patch. These labels are crucial for supervised learning tasks, as they provide the ground truth for the presence or absence of karst features within each patch. The labels are designed to facilitate the training of machine learning models that can automatically detect and classify karst phenomena from the spectral data provided.