Terrain Recognition & Ground-Truth Lab

LiDAR terrain fingerprints + 60 geotagged field photos + field-useful feature classifications.

Prototype status

This is an explainable similarity engine, not a trained mine detector. The 60 photos are ground-truth observations. Record the physical feature that actually exists at each location so the similarity engine learns from real mine workings, supporting evidence, natural terrain, and failed searches.

Reference library

60photo observations

0reviewed labels

3LiDAR-driven hikes

03DEP fingerprints cached

Offline route and image fingerprints are ready.

Select a photograph

Tap a photo marker to open its photograph and field classification here.

Show dataset

Closest reference photographs

Select an observation or candidate. Similarity initially uses route, mining context, and visual fingerprints. Building 3DEP fingerprints adds relief, slope, curvature, depression, and roughness.

Prediction candidates

Click a target to build a live 3DEP fingerprint and compare it with the photographed terrain.

Scan map center

Samples a 5 × 5 grid across roughly 800 meters and ranks the cells by similarity to your reviewed positive examples. With no labels, it ranks against the entire photo library.

Field way-finding card

Choose a target, scan result, or photograph. Use GPS to calculate range and bearing.

● photos◇ targets1–8 scan rank