recipe_rex

Project for CS329s: Machine Learning System Design

https://stanford-cs329s.github.io/reports/tender-recipe-recommendations/

Download Requirements

Run pip install -r requirements.txt to install python dependencies

Scrape recipes from webiste

Raw data is scraped by running the scrape_recipes.ipynb notebook in order to scrape Asian websites for around 5 websites

Saving Features and Data

0. cd feature_generation

1. Run python save_recipes.py in order to save the clean_data/all_recipes_nutrient.csv and clean_data/url_index_mapping.csv csv files

2. Run python save_images.py in order save the image files from the scraped websites locally to the images directory.

3. Run python center_crop_resize_images.py in order to resize and center crop all images to 200 by 200 to the images_resized directory so that thumbnails are all standardized

4. Run python get_embeddings.py in order to save the embeddings needed to run recommendation algorithms. This saves 4 files: clean_data/{folder}/nutrient_features.csv, clean_data/{folder}ing_features.csv, clean_data/{folder}embeddings_features_mvp.csv, and clean_data/embeddings_mvp_scaled.csv, where the embeddings scaled is used for recipe similarity and the other csv's are intermediate.

-- Folder represents ['mains', 'desserts', 'sides']

5. Run python label_recipes.py to get the keyword labels for our recipes

Serving Recommendations

1. serve_recs.py contains relevant API class for grabbing urls and images in order to serve and sample recipes for application

2. streamlit run ucb_rex.py for ucb algorithm app locally