Advanced Configuration Options

This reference guide details the advanced configuration options available in WrenchML for fine-tuning your synthetic datasets and optimizing them for specific computer vision tasks.

Advanced CV Model Settings

The Advanced CV Model Settings section allows you to optimize your datasets for specific computer vision applications and deployment targets.

Task Type

The Task Type setting configures the dataset for different computer vision approaches:

Object Detection

• Purpose: Finding and localizing defects within an image
• Annotations: Bounding boxes around defects with class labels
• Best For:
  • Identifying the presence and location of defects
  • Multiple defect detection in a single image
  • Applications where precise defect boundaries aren't critical
• Common Frameworks: YOLO, Faster R-CNN, SSD, EfficientDet
• Output Format: COCO JSON with bounding box annotations

Segmentation

• Purpose: Pixel-perfect identification of defect areas
• Annotations: Pixel masks defining exact defect boundaries
• Best For:
  • Precise defect boundary detection
  • Measuring defect size and shape accurately
  • Applications requiring detailed defect characterization
• Common Frameworks: Mask R-CNN, U-Net, DeepLab, SegNet
• Output Format: COCO JSON with segmentation mask annotations [Content continues in same pattern with proper Notion formatting for all sections...]