New — the Pakistan PV Atlas is live: 18,827 MWp best estimate, mapped cell by cell. Open the atlas →
logo
earthpv
EarthPV
Initializing search
    open-energy-transition/earthpv
    • Overview
    • PV Atlas
    • How it works
    • Setup
    • Experiments
    • Open questions
    • About
    open-energy-transition/earthpv
    • Overview
      • EarthPV
      • How it works
      • Detection model
      • Segmentation & the building map
      • Rooftop classifier (roofclf)
      • Calibration
      • Validation against MaStR
      • Calibration quadrats overview
      • Quadrat protocol
      • Roofclf random-cell validation
      • Solar glint
      • Glint image gallery
      • Validity and limitations
      • Tutorial
      • Setup New Country
      • Contribute a calibration area
      • Experiments: what was tried, and what it cost
      • Calibration boxes log
      • Quadrats as training data
      • Roofclf national deployment
      • Roofclf cell-edge false positives
      • Quadrat-supervised fraction retrain
      • Fraction-head hard-negative retrain
      • SPPI spectral index
      • Boom-window stacking
      • Standard pose matched filter
      • Glint spike-rate estimator
      • Glint PSF and matched filtering
      • Glint tile-batched coverage
      • Glint-validated training labels
      • Epoch jump as recall signal
      • Calibration imagery dating
      • Small ground-mount instrument
      • Calibration density mismatch
    • Open questions
      • TraceTheSun concept note

    EarthPV

    Previous
    Overview
    Next
    How it works
    Open Energy Transition. Code under MIT, documentation and data under open licences.
    Made with Material for MkDocs