AI Models Can Now Identify Individual Trees from Forest Scans

This study investigates how deep learning models can generalize across different forest environments for segmenting individual trees from point cloud data. By training one model on seven diverse datasets—including various laser scanning methods and forest types—the researchers reveal that models can adapt from sparse coniferous data to dense deciduous forests, but struggle in reverse. Their findings highlight the importance of diverse, labeled forest data for creating robust AI systems capable of improving global forest monitoring and carbon sequestration analysis.