Sitemap

Building Simplification refinement and back to a single option

2 min readMar 20, 2026

--

The #Mapflow #Buildings simplification should remain a single option. We reverted the UX update we introduced in December 2025.

Back then, we tried splitting into two options for the “🏠 Buildings”:

  • regularisation” (“BUILDINGREGULARISER” — optimized for right-angled structures, producing cleaner and more orthogonal geometries)
  • simplification” (“DYNAMIC_GRID” — preserved original shapes more closely, better for curved or irregular buildings)

Btw, one can find the “simplification confidence score” in the building feature attributes — usually, it’s higher when it’s the “dynamic_grid”, because of the better shape approximation to the original pixel mask.

Press enter or click to view image in full size
Feature properties (simplification_confidence_score)

Since then, we have learned that actually most of the users who used “🏠 Buildings” chose both “regularisation” and “simplification” (4,960), and most of the rest of the users used the “regularisation” option (1,473) only, while just a small fraction chose the “simplification” only (165).

Based on the recennt history — the distribution of the options chosen for the 🏠 Buildings model

👉 In other words, the vast majority of users either combined both approaches or prioritized the mapping look.

So, based on the feedback from our commercial and free users, we decided to refine ‘‘simplification’’ and make it a (great 🏆) single, simple option again — yet adopted for different building geometries (rectilinear + curved) in the different construction patterns.

The simplification is enabled by default. And using only one option makes it cheaper for users.

See the examples across different urban patterns:

Press enter or click to view image in full size
Urban area, mixed, regular shapes. (Some shapes are intersected and replaced with OpenStreetMap polygons — additional option. Countours are shifted to the footprints based on building height estimates)
Press enter or click to view image in full size
High-dense urban area (biased to simple rectangular shapes due to the lower simplification confidence score)
Press enter or click to view image in full size
University campus area, complex shapes (the mix of methods worked out)
Press enter or click to view image in full size
Urban area mixed, complex shapes (biased to buildingregulariser due the complex shapes with right angles)

BTW, as you process Buildings or Vegetation, take a moment to rate the results ⭐ — it helps us refine the models and improve outcomes for the entire mapping community. 🙏

--

--

GeoAlert
GeoAlert

Written by GeoAlert

We apply Machine learning to automated analysis over Earth observation data