Computer Vision based Automated Timber Structural Defect Detection Framework

Abstract

In the wood construction industry, timber structural defect detection is usually considered a premanufacturing inspection step done manually. To address this issue, the proposed study discusses the timber structural defect detection method based on YOLOVS variants. The evaluation matrices used are precision, recall, mAP.5, and mAP.5-.95, and the results indicate stable convergence and consistent accuracy on the complex dataset instances. This research contributes to the automation of timber defect detection for precise and robust manufacturing of timber structures. The proposed method further improves resource utilization and contributes towards eliminating waste in the residential construction industry.