LiDAR model improves tree species recognition with cross-source limits
Three LiDARs, Twenty-Five Datasets: Cross-Source Generalization in Tree-Species Recognition
Computer Vision and Pattern RecognitionArtificial IntelligenceGraphics
Summary
Identifying tree species from different types of LiDAR scans is tricky because data from different devices and sources vary a lot. The authors collected a large dataset combining 25 different LiDAR sources and built a new model that recognizes tree species better overall. However, the model performs well mostly when tested on data similar to its training sources and struggles to generalize to new, unseen sources. This finding suggests that future tree species recognition efforts should use more balanced tests that consider these source differences.
What this means in practice
- •For forest management teams: Improve automated tree species identification using LiDAR data combined from multiple collection methods to support forest health monitoring.
- •For environmental mapping companies: Develop vegetation mapping services that account for variability across different LiDAR sensors by using multi-source trained models.$Commercial implications: Enables reliable tree species mapping products adaptable to diverse LiDAR data inputs for customers in forestry and ecology.
Authors
Jae Joong Lee, Bedrich Benes
Abstract
FOR-species20K is the largest public LiDAR benchmark for tree-species classification, pooling 25 independently collected terrestrial, mobile, and airborne laser-scanning datasets across 33 species. We introduce a multi-resolution, multi-view ConvNeXtV2 model that improves official hidden-test accuracy from 79.5% to 85.0%, ranking first at the time of submission. We then examine what this score measures. Because the hidden test is species-balanced, its overall accuracy is more comparable to a macro metric than to accuracy under the public set's long-tailed distribution. Moreover, the official random split places every source dataset in both training and evaluation. A logistic probe recovers source identity from out-of-fold model probabilities with 69% accuracy, while geometric features identify the source above 90% within two widely shared species; 18 of the 33 species occur in only one source. Under a source-species-cell holdout, evaluated on 5,566 trees from eight eligible species that retain the same LiDAR-type training support in other sources, the unweighted species mean falls from 82% to 36%, and tree-weighted accuracy falls from 95% to 60%. Count-matched retraining and experiments with two alternative backbones indicate that this gap is not explained by reduced training support alone, while an external PureForest experiment shows a smaller but consistent grouped-split effect. These results establish the 85% leaderboard score as strong in-distribution performance, but not as an estimate of source-disjoint generalization, motivating balanced metrics and source-aware evaluation. The code is available at: https://github.com/JaeLee18/FOR20K-lidar-tree-species