🏠 Buildings v.2026-07-06 — per-location benchmark

This page details the validation of the 🏠 Buildings v.2026-07-06 segmentation model (Global domain, 0.3 m / z19) on 14 areas of interest (AOI), compared against the previous version v.2025-12-10 (the current production 🏠 Buildings model). For each AOI the two prediction masks are shown side by side; click any image to open it full size, and use the ← / → arrow keys to browse between them.

All metrics are area-based: IoU is the intersection-over-union of the predicted and ground-truth building masks, and F1 / Precision / Recall are computed on the overlapping mask area. Evaluation runs: 2026-06-13 (global set) and 2026-06-16 (satellite set).

← → to browse · ESC to close

Mask colour legend: v.2026-07-06 · v.2025-12-10

Aerial imagery validation set

Validation on the global set of 9 areas of interest (mixed urban / suburban / rural).

United States — Fort Myers

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.907	0.951	0.983	0.921
v.2025-12-10	0.880	0.936	0.974	0.901

v.2025-12-10 — United States — Fort Myers — v.2025-12-10

Canada — Rigaud

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.870	0.930	0.934	0.927
v.2025-12-10	0.842	0.914	0.923	0.907

v.2025-12-10 — Canada — Rigaud — v.2025-12-10

South Africa — Worcester

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.592	0.744	0.867	0.651
v.2025-12-10	0.385	0.556	0.845	0.415

v.2025-12-10 — South Africa — Worcester — v.2025-12-10

New Zealand — Wellington

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.838	0.912	0.900	0.924
v.2025-12-10	0.790	0.883	0.888	0.878

v.2025-12-10 — New Zealand — Wellington — v.2025-12-10

Côte d’Ivoire — Bangolo

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.663	0.797	0.844	0.755
v.2025-12-10	0.642	0.782	0.927	0.676

v.2025-12-10 — Côte d'Ivoire — Bangolo — v.2025-12-10

United Kingdom — London

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.813	0.897	0.873	0.922
v.2025-12-10	0.730	0.844	0.800	0.892

v.2025-12-10 — United Kingdom — London — v.2025-12-10

Australia — Adelaide

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.868	0.929	0.914	0.945
v.2025-12-10	0.830	0.907	0.899	0.915

v.2025-12-10 — Australia — Adelaide — v.2025-12-10

United States — Phoenix

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.903	0.949	0.944	0.953
v.2025-12-10	0.847	0.917	0.902	0.933

v.2025-12-10 — United States — Phoenix — v.2025-12-10

New Zealand — Lower Hutt

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.867	0.929	0.956	0.903
v.2025-12-10	0.814	0.898	0.931	0.867

v.2025-12-10 — New Zealand — Lower Hutt — v.2025-12-10

Satellite imagery validation set

Validation on the global set of satellite imagery across 5 dense urban areas.

United Arab Emirates — Abu Dhabi

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.751	0.858	0.828	0.890
v.2025-12-10	0.717	0.835	0.817	0.854

v.2025-12-10 — United Arab Emirates — Abu Dhabi — v.2025-12-10

India — Bangalore

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.779	0.876	0.859	0.894
v.2025-12-10	0.735	0.847	0.848	0.846

v.2025-12-10 — India — Bangalore — v.2025-12-10

Saudi Arabia — Riyadh

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.817	0.899	0.920	0.879
v.2025-12-10	0.654	0.791	0.687	0.931

v.2025-12-10 — Saudi Arabia — Riyadh — v.2025-12-10

India — Thane

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.768	0.869	0.889	0.849
v.2025-12-10	0.722	0.839	0.871	0.809

v.2025-12-10 — India — Thane — v.2025-12-10

Russia — Ufa

Model	IoU	F1	Precision	Recall
v.2026-07-06	0.821	0.902	0.897	0.907
v.2025-12-10	0.818	0.900	0.897	0.903

v.2025-12-10 — Russia — Ufa — v.2025-12-10

Summary

v.2026-07-06 leads the previous production model on F1 in all 14 AOIs. On the global validation set of aerial imagery (9 AOIs) the mean area-based F1 rises from 0.849 to 0.893 (IoU 0.751 → 0.813); on the 5 dense-urban satellite imagery AOIs the mean F1 rises from 0.842 to 0.881 (IoU 0.729 → 0.787), driven mainly by higher recall and precision in informal and high-density built-up areas such as Riyadh, Bangalore and Thane.