Casting Defect Inspection System, Raghav Singh

Visual inspection of cast parts is slow and inconsistent between operators. I built a system that classifies a casting from a single image and returns a severity score and an accept or reject call, with a saliency overlay so an inspector can see what the model reacted to rather than taking the label on trust.

Model data
Task	Binary classification, defective or acceptable
Dataset	7,000+ labeled cast-metal images (public)
Backbone	ResNet18, ImageNet pretrained
Training	Two-stage: frozen backbone, then unfrozen final block
Test accuracy	96%
Outputs	Class, confidence, severity score, accept/reject
Stack	Python · PyTorch · Streamlit

Engineering decisions

Transfer learning over training from scratch. 7,000 images is far too few to learn general edge and texture features, so the pretrained backbone supplies them and only the task-specific layers are fitted.
Two-stage fine-tuning. Training the head first with the backbone frozen prevents large early gradients from wrecking the pretrained weights. Unfreezing the final block afterwards adapts the highest-level features to cast surfaces.
Grad-CAM as the acceptance check. Accuracy alone doesn't show whether the model learned the defect or a lighting artifact. The heatmap confirms activation sits on the flaw.
Severity score, not just a label. A binary pass/fail gives an inspector nowhere to triage. Grading the output lets borderline parts be routed for review rather than scrapped outright.