A comprehensive benchmark of multi-generation YOLO architectures for forest species identification from macroscopic wood images
A comparative study of 15 models from the YOLOv8, YOLO11, and YOLO26 families, in five sizes each, for automated classification of 73 forest species from macroscopic wood images. The processed dataset, trained models, and interactive tools are available here.
Spancerski, J.S., de Paula Filho, P.L., Kugler, M. et al. A comprehensive benchmark of multi-generation YOLO architectures for forest species identification from macroscopic wood images. Sci Rep 16, 25802 (2026). https://doi.org/10.1038/s41598-026-54925-y
Why automate identification?
Illegal logging
Enforcement relies on rapidly identifying seized wood species — a process that currently demands scarce specialists.
Shortage of anatomists
Few professionals master wood anatomy. Samples are sent via courier for remote analysis, causing delays and inaccuracies.
Field-ready solution
A model deployed on a smartphone allows environmental agents to identify species on-site, generating instant reports.
Datasets and models
Combined datasets
15 models compared
All trained with mosaic augmentation · 448×448 px · 70/15/15 split
Model performance
Four models reached 99.87% test accuracy (11 misclassifications out of 8,640 patches), and all 15 reached 100% top-3 accuracy. YOLO26m offered the best accuracy–efficiency trade-off. Detailed separability and misclassification analysis is available in the t-SNE Viewer.
YOLO26m
YOLO26 family · medium variant · best accuracy–efficiency trade-off
Explore data and models
Paper
A comprehensive benchmark of multi-generation YOLO architectures for forest species identification from macroscopic wood images
Spancerski, J.S., de Paula Filho, P.L., Kugler, M. et al. A comprehensive benchmark of multi-generation YOLO architectures for forest species identification from macroscopic wood images. Sci Rep 16, 25802 (2026). https://doi.org/10.1038/s41598-026-54925-y