README.md

metadata

title: Medical Image AI Lab
emoji: 🔬
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false

🔬 Medical Image AI Lab

An educational demo for ML/AI students, researchers, and educators

Learn how computer vision models analyze and misclassify real dermoscopy images.

🎓 Educational Purpose

This interactive demo lets you explore:

• How ML models handle ambiguous medical images
• The difference between confidence and correctness
• Why medical AI is challenging
• Dataset bias and class imbalance effects
• Model uncertainty and calibration

📊 Model Details

• Architecture: Vision Transformer (ViT) with BiomedCLIP weights
• Dataset: HAM10000 (10,015 dermoscopy images)
• Test Accuracy: 51.16% (3.6x better than random guessing)

⚠️ Disclaimer

For educational and research purposes ONLY. NOT for medical diagnosis.

Always consult a board-certified dermatologist for actual medical concerns.