How solfazzer/derma-ai works
Derma-AI is an academic prototype for dermoscopic-image classification and Grad-CAM explanations, but the checked-in runtime is currently only a FastAPI health endpoint. The documented broader workflow includes a web interface, model training, and local model use; those capabilities are represented as planned or documented components, not as implemented request flow. The API reads settings from environment-backed configuration. Edges for training and model handling follow README statements; unsampled implementation details are intentionally left unconnected.
Main entry points: main.py FastAPI service · app Image classification · frontend Web interface · models Local model file
Components 11
API
- FastAPI service
app/main.py - Health endpoint
app/main.py - Runtime settings
app/config.py
Prediction product
- Web interface
frontend - Image classification
app - Grad-CAM explanation
app
Model lifecycle
- HAM10000 training
notebooks/01_entrenamiento_ham10000.ipynb - Local model file
models - Local data and SQLite
data
Other components
- User
- Google Drive
Connections 9
- User → FastAPI service: requests health
README.md - FastAPI service → Health endpoint: routes health
app/main.py - FastAPI service → Runtime settings: reads settings
app/main.py - User → Web interface: uses interface
README.md - Web interface → Image classification: submits images
- Image classification → Grad-CAM explanation: explains prediction
README.md - HAM10000 training → Google Drive: stores model
README.md - Google Drive → Local model file: copies model
README.md - Local model file → Image classification: provides model
README.md