How to use posters analyser

Code Explenation

The code is separated into 5 notebooks, each has its own responsibility but is not necessarily independent. The notebooks are:

• posters_fetcher
• duplicate_remover
• actor_fetcher
• face_recognition_ethnicity
• analysis

Below you can see a quick explanation of each notebook, they need to be run in that order.

Posters_fetcher - Posters fetching

• Download metadata from imdb
• Download posters images
• Build movie_posters dataframe
• Save tmdb_data

Duplicate_remover - Duplicates removal

• Load movie_posters
• Calculate each image hash value
• Assign a “dup” value of True or False to each poster
• Save to posters_with_dup dataframe

Actor_fetcher - Actors fetching

• Load movies ids from posters_with_dup
• Save movies and actors metadata
• Save actors images
• Create actors dataframe and save as actors_df

Face_recognition_ethnicity - Face & Ethnicity recognition

• Load posters_with_dup dataframe
• Perform posters detection and save as posters_face_encodings
• Perform posters encoding and save as posters_face_encodings
• Load actors_df dataframe
• Perform actors detection & encoding and save as actors_face_encodings
• Recognize the actors who appear in the posters and save as match_poster_actor_cast_all
• Use fairface 4 races model to predict each actors ethnicity
• Add ethnicity information to the posters dataframe and save as posters_new_races4_cast_all
• Create ranking dataframe to save the information about the actors positions in the cast list, saved as ranking_posters_new_races4_cast_all

Analysis

• Data preperation
• Graphs creation seperated by titles

Demo

There is a demo you can run to test the code, we uploaded the data and the dataframes you should expect to get.

Stages of the demo:

• In poster_fetcher notebook, use sample_aids. The demo is the default so comment those lines for a full run.
• In duplicate_remover notebook, load posters sample zip and previous obtained dataframes.
• In actor_fetcher notebook use previous obtained dataframes
• In Face_recognition_ethnicity notebook, load actos zip, movies zip, posters zip, w600k_r50.onnx model and use previous obtained dataframes
• In analysis use previous obtained dataframes