Wood knots defects dataset

This dataset contains high-resolution annotated images of wooden glulam beams, specifically focused on knots and surface defects. It includes two subsets of images captured at different distances (100 cm and 70 cm) to evaluate segmentation performance under varying scales. Each image is accompanied by a manually created ground truth segmentation mask. The dataset was used to fine-tune and evaluate deep learning models for semantic segmentation tasks, particularly for research on the domain adaptation and generalization capabilities of the Segment Anything Model (SAM) in real-world, non-synthetic wood inspection contexts. The dataset is intended to support further research in computer vision, quality control, and digital fabrication applications in the wood industry.