Multiple datasets of PCB images standardized into the COCO JSON format for defect detection tasks. It provides annotated images identifying manufacturing flaws across various printed circuit board designs.
Use Cases
- Train object detection models like YOLO or Faster R-CNN using the bbox and category_id fields to identify manufacturing errors.
- Evaluate model generalization across different PCB types by leveraging the standardized images metadata.
- Perform error analysis on specific defect types by filtering the annotations list by category_id.
Strengths
- Standardized COCO JSON format including images, annotations, and categories keys.
- Includes bounding box coordinates (bbox) and category IDs for various PCB manufacturing defects.
- Aggregates multiple source datasets into a single consistent schema for cross-dataset evaluation.