stonebystone.io

The Technology

The stmap Lifecycle

From raw orthophotos to georeferenced stone-by-stone analysis. Our pipeline combines state-of-the-art computer vision with specialized domain knowledge in masonry.

1

Input: High-Resolution Imaging

We process ultra-high-resolution orthophotos (RGB, Ambient Occlusion, and Normal maps). These provide the raw geometric and visual information needed for accurate stone detection.

  • Support for Multi-GB GeoTIFFs
  • Sub-millimeter Ground Sampling Distance (GSD)
Input Data
ML Training
2

Training: Domain Adaptation

Our models are trained on a diverse dataset of over 24 masonry sources, including sandstone, brick, and natural stone. We use advanced augmentation and active learning loops to ensure robustness.

Epoch 42/100 | IoU: 0.9562 | Loss: 0.0431
3

Output: Precise Vectorization

The final output is not just a mask, but individual georeferenced polygons. Each stone is identified as a unique instance, ready for damage assessment and monitoring.

DXF GeoJSON GPKG
Vector Output

Ready to map your masonry?

Join our pilot program and get automated stone analysis for your restoration projects.

Get Started