How hemanthsai456/leveraging-poster-and-trailer-audio-embeddings-for-cold-start-movie-recommendation works
This is a notebook-based research movie recommender with a Streamlit interface. Users can request personalized Top-N results, filter by genre, get recommendations related to a selected movie, or predict a rating for a selected movie. The inference wrapper loads saved models and aligned catalog features, excludes movies already seen, and routes candidates through the regular or cold-start model. README-described research and evaluation stages are included, but the notebook itself was not sampled, so its internal relationships are not detailed.
Main entry points: recommender.py Inference wrapper · app.py Streamlit interface · 01_Notebook.ipynb Data and model research · recommended_artifacts Models and aligned catalog
Components 14
Research pipeline
- Data and model research
notebooks/01_Notebook.ipynb - Model evaluation
notebooks/01_Notebook.ipynb
Recommendation experience
- Streamlit interface
app.py - Movie search
app.py - Top-N and genre requests
app.py - Movie-based requests
app.py - Rating requests
app.py
Inference engine
- Inference wrapper
recommender.py - Candidate filtering
recommender.py - Known-movie scoring
recommender.py - Cold-start scoring
recommender.py - Similarity and ranking
recommender.py
Saved assets
- Models and aligned catalog
recommended_artifacts
Other components
- Movie fan
Connections 17
- Movie fan → Streamlit interface: submits requests
README.md - Streamlit interface → Inference wrapper: calls inference
app.py - Streamlit interface → Movie search: searches titles
app.py - Streamlit interface → Top-N and genre requests: requests recommendations
app.py - Streamlit interface → Movie-based requests: requests similar picks
app.py - Streamlit interface → Rating requests: requests rating
app.py - Movie search → Inference wrapper: calls search
app.py - Top-N and genre requests → Inference wrapper: calls recommend
app.py - Movie-based requests → Inference wrapper: calls hybrid recommendations
app.py - Rating requests → Inference wrapper: calls prediction
app.py - Inference wrapper → Models and aligned catalog: loads models and features
recommender.py - Inference wrapper → Candidate filtering: filters candidates
recommender.py - Inference wrapper → Known-movie scoring: scores known movies
recommender.py - Inference wrapper → Cold-start scoring: scores unseen movies
recommender.py - Inference wrapper → Similarity and ranking: combines and ranks
recommender.py - Data and model research → Model evaluation: evaluates models
README.md - Data and model research → Models and aligned catalog: produces saved artifacts
README.md