How jishnudipsaha/eye-disease-prediction works
This repository provides OCT retinal-image classification through a FastAPI web app, with a legacy Streamlit interface, class-specific recommendation text, and a training notebook/model artifacts. The sampled FastAPI source shows upload preprocessing, model inference, class/confidence selection, and recommendation lookup. The frontend file and notebooks were not sampled, so their internal behavior and wiring are not detailed.
Main entry points: app.py FastAPI app · app_streamlit.py Streamlit app · index.html Web interface · Trained_Model.h5 Trained weights
Components 11
Web applications
- FastAPI app
app.py - Web interface
static/index.html - Streamlit app
app_streamlit.py
Model lifecycle
- Image preprocessing
app.py - Classifier model
app.py - Trained weights
Trained_Model.h5 - Keras model
Trained_Model.keras - Model training
Training_model.ipynb - Inference notebook
Model_Prediction.ipynb
Clinical guidance
- Class recommendations
recommendation.py
Other components
- User
Connections 12
- User → Web interface: opens
- User → FastAPI app: uploads scan
- FastAPI app → Web interface: serves page
- FastAPI app → Image preprocessing: preprocesses image
- Image preprocessing → Classifier model: provides tensor
- Classifier model → Trained weights: loads weights
- Classifier model → Class recommendations: selects class guidance
- Model training → Trained weights: trains weights
- Inference notebook → Keras model: uses model
- User → Streamlit app: uses legacy app
- Streamlit app → Keras model: loads model
- Streamlit app → Class recommendations: displays guidance