This large-scale image dataset features wood surface defects captured at a consistent resolution of 2800 x 1024 pixels for industrial quality control. It provides dual-layer annotations including bounding boxes in YOLO format and semantic maps for precise defect localization.
Use Cases
- Train object detection models to identify wood flaws using the YOLO-formatted bounding boxes
- Develop semantic segmentation architectures for pixel-level defect delineation using the semantic maps
- Benchmark the performance of vision-based quality control algorithms on high-aspect-ratio 2800 x 1024 pixel imagery
Strengths
- All images are standardized to a high-resolution format of 2800 x 1024 pixels
- Includes dual annotation types: YOLO-formatted bounding boxes and semantic maps
- Images are optimized using PIL with a specific JPEG quality setting of 50 for storage efficiency