Explainable PCB Defect Classifier

Upload a cropped PCB defect region, choose a model, and get a real-time prediction with a Grad-CAM explanation of which pixels drove the decision.

Select Model

ResNet50 (98.43% accuracy) MobileNetV2 (96.86% accuracy) Custom CNN (88.12% accuracy)

Upload PCB Defect Crop

Try a sample defect

Prediction Confidence

Grad-CAM Explanation