MillionTrees: A global benchmark for tree detection across geography, sensor, and annotation geometry
- 1University of Florida
Abstract
Airborne tree monitoring using remotely sensed data, combined with the growth of computer vision models for object detection and segmentation, can facilitate broad scale assessment of tree health, ecology and ecosystem function. However, datasets for airborne tree detection have been largely limited to a single geography, resolution, or annotation approach, making it difficult to develop general models and benchmark their utility for broad applications. We compile existing and newly published annotation datasets to create a comprehensive dataset for training and evaluating airborne tree detection in RGB imagery. The MillionTrees dataset has over a million individual trees from a broad range of forest types, tree species, and geographic locations. In addition, it has six million weakly supervised annotations from a range of sources including airborne LiDAR and photogrammetry, arborist surveys, and simulated landscapes. An evaluation standard is provided to facilitate rapid model development. We benchmark several existing tree detection algorithms and demonstrate that while current open-source models do not generalize across the diverse datasets in the benchmark, training on MillionTrees improves both within-domain and out-of-domain performance. Held-out validation datasets using ground truth derived from terrestrial LiDAR paired with overhead RGB show that overfitting to test data is common and hyperparameter tuning is an underappreciated challenge towards a global zero shot tree detection method. We discuss the limitations of the currently available data for airborne RGB tree detection, how to create more robust evaluation data and the need for better understanding of the uncertainty and utility of tree detection from airborne remote sensing.
Related
- Where single-cell foundation models help: a benchmark of annotation, perturbation and regulatory inference against simple baselines
- An annotation-overlap-flagged rare-disease gene-prioritisation benchmark and PMC index recipe
- Cell-type separability predicts annotation accuracy and outweighs algorithm choice: a factorial benchmark across seven scRNA paradigms
- NED-Net: an annotation-training loop for rodent EEG event detection
- A Benchmark for Gaze-Based Target Detection in Immersive Virtual Reality: Effects of Eccentricity, Size, and Speed on Observation and Detection Dynamics
BibTeX
@misc{weinstein2026milliontrees,
title = {{MillionTrees: A global benchmark for tree detection across geography, sensor, and annotation geometry}},
author = {B. G. Weinstein and J. Veitch-Michaelis and M. J. Allen and A. Altamirano and S. Anacleto and R. Astrup and J. G. C. Ball and H. Baudchon and S. Beery and S. A. Bohlman and J. Brandt and M. Cloutier and S.-O. Duguay and J. Ertel and Z. Feng and A. Firoze and J. Frey and G. A. Fricker and D. Gominski and S. Graves and S. W. D. Grieve and S. de Haas and S. H. M. Hickman and T. D. Jackson and A. Jansen and T. Kattenborn and T. Khan and R. Kwon and K. K. Htoo and E. Laliberte and I. Lefebvre and E. R. Lines and M. Lucas and A. Miranda and B. Mola-Yudego and M. Mugabowindekwe and J. Nicholson and M. Onishi and Y. Onoda and A. Ouaknine and H. J. F. Owen and C. Pawlak and S. Puliti},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.30.762845},
url = {https://www.biorxiv.org/content/10.64898/2026.10.30.762845v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - MillionTrees: A global benchmark for tree detection across geography, sensor, and annotation geometry
AU - B. G. Weinstein
AU - J. Veitch-Michaelis
AU - M. J. Allen
AU - A. Altamirano
AU - S. Anacleto
AU - R. Astrup
AU - J. G. C. Ball
AU - H. Baudchon
AU - S. Beery
AU - S. A. Bohlman
AU - J. Brandt
AU - M. Cloutier
AU - S.-O. Duguay
AU - J. Ertel
AU - Z. Feng
AU - A. Firoze
AU - J. Frey
AU - G. A. Fricker
AU - D. Gominski
AU - S. Graves
AU - S. W. D. Grieve
AU - S. de Haas
AU - S. H. M. Hickman
AU - T. D. Jackson
AU - A. Jansen
AU - T. Kattenborn
AU - T. Khan
AU - R. Kwon
AU - K. K. Htoo
AU - E. Laliberte
AU - I. Lefebvre
AU - E. R. Lines
AU - M. Lucas
AU - A. Miranda
AU - B. Mola-Yudego
AU - M. Mugabowindekwe
AU - J. Nicholson
AU - M. Onishi
AU - Y. Onoda
AU - A. Ouaknine
AU - H. J. F. Owen
AU - C. Pawlak
AU - S. Puliti
PY - 2026
DA - 2026/10/30
PB - bioRxiv
DO - 10.64898/2026.10.30.762845
UR - https://www.biorxiv.org/content/10.64898/2026.10.30.762845v1
AB - Airborne tree monitoring using remotely sensed data, combined with the growth of computer vision models for object detection and segmentation, can facilitate broad scale assessment of tree health, ecology and ecosystem function. However, datasets for airborne tree detection have been largely limited to a single geography, resolution, or annotation approach, making it difficult to develop general models and benchmark their utility for broad applications. We compile existing and newly published annotation datasets to create a comprehensive dataset for training and evaluating airborne tree detection in RGB imagery. The MillionTrees dataset has over a million individual trees from a broad range of forest types, tree species, and geographic locations. In addition, it has six million weakly supervised annotations from a range of sources including airborne LiDAR and photogrammetry, arborist surveys, and simulated landscapes. An evaluation standard is provided to facilitate rapid model development. We benchmark several existing tree detection algorithms and demonstrate that while current open-source models do not generalize across the diverse datasets in the benchmark, training on MillionTrees improves both within-domain and out-of-domain performance. Held-out validation datasets using ground truth derived from terrestrial LiDAR paired with overhead RGB show that overfitting to test data is common and hyperparameter tuning is an underappreciated challenge towards a global zero shot tree detection method. We discuss the limitations of the currently available data for airborne RGB tree detection, how to create more robust evaluation data and the need for better understanding of the uncertainty and utility of tree detection from airborne remote sensing.
ER -