FOR-age: Benchmarking individual tree age estimation using 3D deep learning on dense laser scanning data was just published as a collaboration between SmartForest and the SingleTree project was just published by Stefano Puliti , Binbin Xiang, Marta Vergarechea, Maciej Wielgosz, Terje Gobakken, Eivind Handegard, Juha Hyyppä, Erik Næsset, Nicolas Cattaneo, Mikko Vastaranta, Tuomas Yrttimaa and Rasmus Astrup
Tree age determination of individual trees is crucial for understanding forest dynamics, tree growth, site productivity and describing ecological processes. Traditional methods are invasive, labor-intensive, and costly.
This study investigates the use of deep learning to predict tree age from high-density laser scanning data as a scalable, non-invasive alternative. The dataset includes over 1700 tree point clouds from nearly 1000 trees from Norway, Sweden, and Finland. It includes Norway spruce and Scots pine and a broad range of tree ages, from young seedlings (1 year) to old trees (∼350 years).
The authors approached this as a new downstream task in 3D point cloud deep learning analysis, combining multi-platform laser scanning with lab-based age measurements and state-of-the-art 3D computer vision models.
Highlights
- Tree age predicted from tree point clouds using deep learning methods.
- Dataset includes 1700 trees from boreal forests in Northern Europe.
- Transformer models outperform simpler alternatives for age prediction.
- Pretrained segmentation models used as backbones for regression.
- Models generalize across species and laser scanning platforms.
The paper, code, dataset and benchmarking are available here:
the paper: https://doi.org/10.1016/j.rse.2026.115462 the code: https://github.com/SingleTree-EU/FORage the dataset: https://zenodo.org/uploads/19853987 the benchmarking: https://www.codabench.org/competitions/16014/