Counting Trees from Satellite Imagery with Noisy Supervision
Dimitri Gominski, Maurice Mugabowindekwe , Qiue Xu , Xiaowei Tong, Martin Brandt , Hieu Le , Rasmus Fensholt , Dimitris Samaras , Loic Landrieu · 2026
Read Counting Trees from Satellite Imagery with Noisy Supervision by Dimitri Gominski, Maurice Mugabowindekwe , Qiue Xu , Xiaowei Tong, Martin Brandt , Hieu Le , Rasmus Fensholt , Dimitris Samaras , Loic Landrieu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-24786.txt.
Description
Dimitri Gominski and eight colleagues take on counting individual trees in satellite images. The task matters for environmental monitoring, yet satellite resolution makes it hard. Isolated trees can be spotted, but crown boundaries blur together in dense forest. Hand labelling trees at scale is also far too costly. The team instead derives training labels from airborne LiDAR data. Those labels scale well but arrive noisy, so the counting method must tolerate errors.
Counting is reframed as matching a spatial density, supervised through unbalanced optimal transport. That setup locates lone trees precisely while still estimating density inside dense stands. A self correction step uses transport residuals to clean the noisy labels during training. The new TinyTrees benchmark spans three continents and three satellite sensors. It holds over 216 million tree annotations across some 25,890 square kilometres. The method beats detection, regression, and other transport based baselines throughout, Gominski and colleagues report.
Catalog record
| CREATOR | Dimitri Gominski, Maurice Mugabowindekwe , Qiue Xu , Xiaowei Tong, Martin Brandt , Hieu Le , Rasmus Fensholt , Dimitris Samaras , Loic Landrieu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.24786