Sherpa LLC research demo — normalized scenic:t per segment with a continuous color ramp.
Classes: very_low < 0.15, low < 0.35, medium < 0.55, high < 0.75, epic ≥ 0.75.
Data fusion: Multi-petabyte imagery, global DEM mosaics, land-cover atlases, and nighttime-light priors are staged into aligned tiles so the scenic values stay consistent from alpine passes to coastal highways.
Blockers & fading: National building footprints, infrastructure registries, and activity maps infer massing, heights, and land-use tone, letting the model down-weight factory belts while easing transitions as roads leave towns.
Computation: The HPC core tiles every region, fans rays along each segment from curb-scale 3 km probes out to 15 km vistas, and merges billions of measurements into the per-segment scenic:t, hex colors, and alpha sent to Mapnik.