How yamuna-pinjala/early-detection-of-alzheimer-s-disease-using-variational-autoencoders-on-mri-scans works
This is a Flask web app for Alzheimer’s screening from MRI images, backed by a VAE and latent classifier. Users register or log in, upload an image, pass file and brain-image checks, receive a prediction if confidence is sufficient, and see the result; prediction records are saved to MongoDB. A separate training workflow fits the VAE and classifier. The README gives only a high-level product description; relationships below are based on the shown source excerpts, not unsampled templates or styles.
Main entry points: app.py Web routes · train.py Training workflow · config.py Database access · model.py VAE encoder
Components 17
Web workflow
- Web routes
app.py - Auth view
templates/auth.html - Dashboard
templates/dashboard.html - Result view
templates/result.html - Upload config
config.py
MRI screening
- Image OOD gate
app.py - Pretrained ResNet
- VAE encoder
model.py - Latent classifier
model.py - Class labels
app.py
Persistence
- Database access
config.py - MongoDB
- User records
- Prediction records
Model training
- Training workflow
train.py - ImageFolder data
Other components
- User
Connections 17
- User → Web routes: submits requests
app.py - Web routes → Auth view: renders
app.py - Web routes → Dashboard: renders
app.py - Web routes → Result view: renders
app.py - Web routes → Upload config: checks uploads
app.py - Web routes → Image OOD gate: validates image
app.py - Image OOD gate → Pretrained ResNet: scores image
app.py - Web routes → VAE encoder: encodes MRI
app.py - Web routes → Latent classifier: classifies latent
app.py - Web routes → Class labels: uses labels
app.py - Web routes → Database access: gets database
app.py - Database access → MongoDB: connects
config.py - Web routes → User records: reads and writes
app.py - Web routes → Prediction records: writes prediction
app.py - Training workflow → ImageFolder data: loads images
train.py - Training workflow → VAE encoder: trains encoder
train.py - Training workflow → Latent classifier: trains classifier
train.py