INFO7374-Algorithmic-Digital-Marketing

Folder Information: -

Method 1 :

Contains the .ipynb file of 1st method in the assignment and the csv generated using that method.

Method 2(FAISS Method):

Contains the .ipynb file of the implementaion of facebook AI similarity search method and the csv generated using the same method.

Method 3(Spotify - Approximate Nearest Neighbors Oh Yeah):

Contains both the files to run the spotify annoy method which generates the nearest neighbor json file which can be loaded into elasticsearch using bulk API.

Preprocessing:

Contains preprocessing file which is used to extract images to disk from bson.

Streamlit:

Contains all the images used in both method 1 and 2 along with csv files and py file.

Finale:

Contains all the sampled images which are used for similarity searches.

How to run the similarity searches: -

Method 1: - Use the images from 'finale' folder and run the SimilaritySearchMethod1.ipynb file from 'Method 1' folder.

Method 2 - FAISS: - Use the images from 'finale' folder and run the faissmethod.ipynb file from 'Method 2 - FAISS' folder(Perfferably run on Google Colab)

Method 3 - Spotify: - First run the get_image_features_vectors.py and store the feature vector(.npz) files. After storing the features run the cluster_image_feature_vector.py which will create the similarity indexes and store in nearest_neighbors.json file. Populate the elasticsearch cluster using bulkAPI and run the flask app in 'elasticsearch folder' to display the nearest neighbor in web app.