Stone-wall segmentation (DINO-CV)
Finds historic dry-stone walls in two relief views of a 10 cm lidar DEM.
- Field
- Geomorphology
- Method
- Deep learning
- Data it takes
- imagery
- License
- CC BY-NC 4.0 — restricted use
The model finds historic dry-stone walls in two relief views of a 10 cm lidar DEM, a multi-directional hillshade and a VAT (Visualization for Archaeological Topography) image. It was trained on the Budj Bim Cultural Landscape in Victoria, Australia (Huang et al. 2026, Science of Remote Sensing 13:100372). On held-out test areas the paper reports a mean intersection over union (mIoU) of 68.6%, and 63.8% when trained with 10% of the labels.
Budj Bim is a UNESCO World Heritage site on Gunditjmara country, and the data and models carry the authors’ cultural-sensitivity notice. The license is CC BY-NC 4.0, for research and education only.
Catalog entry last checked 2026-10-11. All models →
Run Stone-wall segmentation (DINO-CV) on your data
The lab runs this on its own compute, in a container, with the inputs and parameters recorded alongside the result. Initial scoping conversations are free.
Contact the Lab