Portfolio-ready medical image classification pipeline

Summary

This release marks the first portfolio-ready version of MedicalImageClassifier.

It includes a reproducible and explainable chest X-ray pneumonia classification workflow with training, evaluation, explainability, inference, testing, CI, and responsible documentation.

Included Features

• PyTorch transfer-learning training pipeline
• CPU smoke-test configuration
• Accuracy, ROC-AUC, PR-AUC, Brier score, and confusion matrix
• ROC and Precision-Recall curve generation
• Grad-CAM examples for TP, TN, FP, and FN cases
• Command-line inference script
• Gradio inference app
• ONNX export CLI
• Unit tests
• GitHub Actions CI
• Model card
• MIT license
• Lightweight multi-agent ML audit workflow

Notes

This project is for educational and portfolio demonstration purposes only. It is not intended for clinical diagnosis or deployment in real healthcare settings.

The trained checkpoint and dataset are not included in the repository.