SolarMap.PH

What can your roof do?

Type your address. We measure your actual roof from satellite and tell you whether solar pays back.

Try:

We'll tell you whether you can install solar on your roof, what it would save you, and what your block's grid saturation means for approval. Address is not stored.

Your average monthly bill. Edit it so the payback is yours, not a stranger's.

Free, no signup. SolarMap.PH runs no backend. The address you type goes straight from your browser to third-party geocoders (Photon, Nominatim, Overpass, Esri tiles, PVGIS) and never reaches a server we control. What gets seen by whom →

Three asymmetries decide whether a Filipino homeowner gets a fair solar deal.

Installer quotes vary 5x for the same system. LGU permit fees range ₱16K-153K for the same project. Net metering caps residential export at 100 kW, targets a 10-working-day utility decision, and depends on a 3-day LGU fast-track that many LGUs have not adopted. Neutral information is hard to come by.

This tool reads your address, finds your building in OpenStreetMap, measures the roof from satellite, runs the payback math at the July 2026 Meralco residential rate of ₱14.8261/kWh, and says what to do next. It also shows nearby rooftops where our computer-vision survey already detected solar, and the local permit cost where we could verify it.

No installer pays to be recommended. No data leaves your browser. The math is open-source and reproducible. How we computed everything →

research backplane

515 rooftops with detected solar across 41 cities

high-confidence
280
per-roof polygons
384
extra regions
7

Detections score above 0.85, calibrated to 96% precision on the held-out NCR validation set. The 7 franchises outside Metro Manila carry their own per-domain calibration or ship as candidate inventory with no precision claim. Survey generated 2026-05-08.

Open the map →

how it works

A frozen CLIP encoder, a calibrated head, then SAM for the panel outlines

How the model was trained, calibrated and validated: the active-learning rounds, the encoder ablation, the per-domain calibration that replaced an over-optimistic cross-region precision, and the segmentation step that turns a tile-level score into a per-building panel polygon.

Read methodology →