Data supporting the publication: Enhancing existing Remote-sensing Datasets with weakly supervised Deep Learning: A Case Study on Antarctic Rock outcrops

Source
DataCite
Published
2026
License
Creative Commons Attribution 4.0 International
Themes
Remote Sensing & Satellites, Permafrost & Glaciers, Machine Learning & AI

Get the data from DataCite

About this dataset

This dataset provides a high-resolution (10 meter) vector polygon mask of continuous rock outcrops along the Antarctic Peninsula (AP). Developed using a weakly supervised deep-learning framework, this mask refines legacy products by combining multi-modal remote sensing and topographic data. It effectively delineates small, fragmented, and heavily shaded outcrops frequently missed by coarser datasets. While this repository currently covers the Antarctic Peninsula—the most rock-diverse region on the continent—it serves as the precursor to an upcoming continent-wide mapping release. For a detailed overview of the framework, methodology, and performance, please refer to the publication.

How the data was collected

Developed using a weakly supervised deep-learning framework, this mask refines legacy products by combining multi-modal remote sensing and topographic data.

Browse sustainability datasets on SusanHub