Published July 10, 2026
| Version v1
Dataset
Open
HiRISE Landform Classification Dataset
Authors/Creators
-
1.
Stanford University School of Medicine
-
2.
National Aeronautics and Space Administration, Jet Propulsion Laboratory
Description
A high-resolution, balanced terrain classification dataset built from RAW HiRISE orbital imagery of the Martian surface. Contains 2,786 image tiles across seven scientifically significant Martian terrain classes (crater, dark dune, slope streak, bright dune, impact ejecta, swiss cheese, spider) at 1024×1024 px resolution (0.5 m/pixel HiRISE source). Includes both raw unbalanced tiles and the balanced train/test split.
The dataset contains a detailed README.md file with all the technical specifications.
Files
terrain-classification-dataset-v.1.0.zip
Additional details
Software
- Repository URL
- https://github.com/pranavdurai10/mantle