How kumarsunny-nitian/fashion-rec-ai works
StyleMatch is a Gradio image-to-image fashion recommender. A user uploads an image; the app preprocesses it, extracts a normalized ResNet50 embedding, compares it with embeddings of images in the local catalog, sorts by similarity, and displays up to three results. The README also describes real-time inference and future filtering, larger catalogs, vector search, and feedback, but those future capabilities are not implemented in the sampled source.
Main entry points: app.py Recommendation handler · images Fashion images
Components 9
User interface
- Gradio interface
app.py - Recommendation gallery
app.py
Recommendation inference
- Recommendation handler
app.py - Image preprocessing
app.py - ResNet50 embedding
app.py - Cosine similarity
app.py - Top-3 ranking
app.py
Image catalog
- Fashion images
images
Other components
- User
Connections 9
- User → Gradio interface: uploads image
- Gradio interface → Recommendation handler: dispatches
- Recommendation handler → Image preprocessing: extracts features
- Image preprocessing → ResNet50 embedding: prepares input
- Recommendation handler → Fashion images: reads candidates
- Recommendation handler → Cosine similarity: compares embeddings
- Cosine similarity → Top-3 ranking: orders scores
- Recommendation handler → Recommendation gallery: returns results
- Recommendation gallery → User: displays items