First-review implementation · 20% scope

Get a mammogram dataset safely into an EfficientNetV2 pipeline.

This starter guide gives you exactly the files and commands needed to prove that local mammogram images can be inspected, preprocessed, augmented, and passed through a frozen pretrained classifier.

Start with dataset inspection View the Python pipeline

Your implementation path

Four small, explainable stages

Inspect dataset

Confirm folders, labels, image counts, and supported formats before changing any code.

Create stable splits

Use a reproducible per-class 70% / 15% / 15% train, validation, and test split.

Prepare images

Decode, resize to 224 × 224, and normalize to the EfficientNetV2 input range.

Verify transfer learning

Freeze EfficientNetV2B0, attach a small softmax head, and run only two epochs.