🌲 Forest and trees

The model is trained on high-resolution data (0.6-0.3m) for different areas and climate zones.

The result includes all areas covered with tree and shrub vegetation, including sparse forest and shrublands.

Model resolution allows to detect small group of trees and narrow tree lines.

The model is robust to region change, and performs well in most environments, including urban. The image should be taken in active vegetation period, because leafless trees or vegetation covered with snow are not the target class.

Latest model tags 🏷️

Version

2026-07-03

Geo Domain

Global

Model method

Segmentation

GSD / Map Zoom

0.6–0.3 m / z18–19

Sample of processing results for solid Forest mask

Additional options:

• Height estimation – forest mask classification by height classes

• Tree crown polygons - extracts tree crowns from forest vegetation as well as free-standing trees, provides them as polygons

• Tree crown points - extracts tree crowns from forest vegetation as well as free-standing trees, provides them as points

Important

We recommend using the Tree crown options with 0.3m resolution imagery (~ 19 zoom) for the best results in case you need to detect individual trees.

Note

Forest Height classification follows the following classes:

• Shrubs lower than 4 meters;

• Forest from 4 to 10 meters high;

• Forest more than 10 meters high;

This classification is used as a decision support for the vegetation management in powerline zones, etc. See the professional solutions by Geoalert. The tresholds can be customized depending on the requirements.

Processing results samples

Sample of results for Tree crowns, points

Sample of results for Forest with heights mask (raster output)

Benchmarks - segmentation

Latest update — 🌲 Forest and trees v.2026-07-03 (Global, Segmentation). The model was evaluated on a validation set of 6 areas of interest (AOI) against manually annotated ground truth. Metrics are area-based: IoU is the intersection-over-union of the predicted and ground-truth vegetation masks, and F1 / Precision / Recall are computed on the overlapping mask area.

AOI (location)	Predicted features	IoU	F1	Precision	Recall
Italy — Crotone	254	0.862	0.926	0.944	0.909
Spain — Madrid	368	0.813	0.897	0.880	0.915
Philippines — Balanga	220	0.806	0.892	0.864	0.923
Spain — Velilla de San Antonio	1070	0.784	0.879	0.894	0.864
Spain — Cuenca	409	0.763	0.865	0.859	0.872
Uzbekistan — Tashkent	156	0.730	0.844	0.905	0.791
Spain — Rus	249	0.617	0.763	0.699	0.841
Argentina — La Banda	93	0.576	0.731	0.653	0.829
Global (mean of 8 AOIs)	2819	0.744	0.850	0.837	0.868

Area-based IoU / F1 / Precision / Recall measured against ground-truth vegetation masks; evaluation run 2026-07-03. Compared with the previous version v.2025-06-14 (mean F1 0.513, IoU 0.396).

See also

📊 See per-location benchmark details for the area-by-area breakdown, including comparison with the previous version and prediction-vs-ground-truth overlays.