AI Just Got Better at Counting Trees

This study evaluates TreeLearn, a deep-learning-based tree segmentation model trained on multi-domain forest point clouds. Results show that fine-tuning the model with both high- and low-resolution datasets (MLS, TLS, UAV) significantly improves instance segmentation performance and generalization across forest types. The findings highlight the importance of diverse, labeled training data to develop AI models capable of accurately mapping trees in varying environments—laying groundwork for scalable, data-driven forest monitoring and management.