Optimizing Lightweight YOLO11 for Wood Defect Detection: A Comparative Analysis of Attention Mechanisms

To address the concurrent demands for precision and efficiency in wood surface defect detection, this study proposes a lightweight YOLO11n framework enhanced with attention mechanisms. We systematically integrated five distinct attention modules into the Backbone and Neck sections of the model, constructing a series of comparative "YOLO11n + Attention" models to investigate the impact of different mechanisms and embedding strategies. Furthermore, a novel three-dimensional weighted evaluation method incorporating accuracy, speed, and resource consumption was proposed for comprehensive model assessment. Experimental results identified the SEAM attention mechanism as the top performer. The YOLO11n + C2PSA_SEAM model achieved the highest detection accuracy, with a mAP@0.5 of 90.4%, a 3.6 percentage-point improvement over the baseline (86.8%). Notably, the YOLO11n + SEAM model ranked first in the comprehensive evaluation, maintaining a high inference speed of 476 FPS and achieving the best computational efficiency. The study also demonstrated that embedding the attention module within the Backbone yields superior overall performance compared to the Neck. This work provides a high-precision, efficient, and practically evaluable lightweight solution for automated defect detection in the wood industry.