Defect Detection & Classification

Image-Based Detection

CNN architectures, SEM images, and wafer map classification

CNNs for Defect Detection

CNNs for Defect Detection

Deep learning, particularly Convolutional Neural Networks (CNNs), has transformed defect detection in semiconductor manufacturing:

  • SEM image classification: After inspection tools locate potential defects, a Review SEM captures high-resolution images. CNNs classify these images into defect categories (particle, bridge, scratch, nuisance, etc.) with >95% accuracy, replacing manual human review.
  • Wafer map pattern recognition: CNNs classify wafer-level defect patterns (center, edge, ring, scratch, random) to identify root causes. Input: 2D defect density map as an image.
  • Object detection: Models like YOLO or Faster R-CNN can locate and classify multiple defects in a single large SEM or optical image.

Common architectures in production:

  • ResNet, EfficientNet: Standard backbone networks for classification
  • U-Net: For segmentation — pixel-level defect delineation
  • Vision Transformers (ViT): Emerging for their ability to capture global context

Key Concept: Data Challenges

Semiconductor defect datasets are notoriously challenging: highly imbalanced (rare defect types), variable image quality, sensitive/proprietary (can't use public pretrained models easily), and expensive to label (requires expert annotators). Data augmentation, few-shot learning, and self-supervised pretraining are active research areas.

Knowledge Check

Knowledge Check

1 / 1

What accuracy level do CNNs typically achieve for SEM defect classification?