deep learning for defect classification

**Deep Learning for Defect Classification** is the **application of CNNs and other deep learning architectures to automatically classify wafer defects from images** — replacing manual defect review with automated, consistent, and faster classification of SEM images, optical inspection images, and wafer maps. **Deep Learning Approaches** - **CNN Classification**: ResNet, EfficientNet trained on defect images to classify defect types. - **Wafer Map Classification**: Classify spatial defect patterns (center, edge, ring, scratch, random). - **Object Detection**: YOLO, Faster R-CNN to localize and classify multiple defects in one image. - **Few-Shot Learning**: Handle new defect types with very few labeled examples. **Why It Matters** - **Consistency**: Eliminates operator-to-operator variability in manual defect classification. - **Speed**: Classifies thousands of defects per second (vs. seconds per defect for manual review). - **Nuisance Filtering**: Automatically separates real defects from nuisance signals (noise, artifacts). **Deep Learning for Defect Classification** is **AI-powered defect review** — using CNNs to automatically classify and sort defects faster and more consistently than human reviewers.