How yamuna-pinjala/fashion-recommendation-system works
This is a Streamlit image-based fashion recommender. A user uploads an image; the app extracts a normalized ResNet50 feature vector, queries a nearest-neighbor index built from saved image features, and displays matching catalog images. The sampled app.py shows this workflow and its runtime dependencies, but the referenced pickle files are not present in the repository tree.
Main entry points: app.py Upload and display
Components 7
User interface
- Upload and display
app.py
Image features
- Feature extraction
app.py - ResNet50 model
app.py
Recommendation retrieval
- Saved image features
- Image filenames
- Nearest-neighbor index
app.py
Other components
- User
Connections 11
- User → Upload and display: uploads image
app.py - Upload and display → Feature extraction: extracts features
app.py - Feature extraction → ResNet50 model: predicts features
app.py - Upload and display → Saved image features: loads features
app.py - Upload and display → Image filenames: loads filenames
app.py - Upload and display → Nearest-neighbor index: fits index
app.py - Saved image features → Nearest-neighbor index: provides vectors
app.py - Upload and display → Nearest-neighbor index: queries neighbors
app.py - Nearest-neighbor index → Upload and display: returns matches
app.py - Upload and display → Image filenames: resolves matches
app.py - Upload and display → User: displays images
app.py