What celebrity is your lookalike?

We have an instance of the application deployed on DigitalOcean.

For details look at Model and Index READMEs.

• Dataset (must be unzipped)
• Dump
• Landmarks

To build and run the application locally:

• Load all needed resources
• Specify paths to resources in environment variables
  • Linux\MacOS: PATH_TO_DATASET=<path to folder with dataset> PATH_TO_DUMP=<path to dump file> PATH_TO_LANDMARKS=<path to landmarks file>
  • Windows: set PATH_TO_DATASET=<path to folder with dataset> & set PATH_TO_DUMP=<path to dump file> & set PATH_TO_LANDMARKS=<path to landmarks file>
• Install Docker and use docker-compose:
  • Build with prepared images: docker-compose up
  • Build from source files: docker-compose -f docker-compose-dev.yml up

• Darya Rovdo

  • Creating network model and Python service for using this model.
  • Dataset preprocessing and model training
  • Making benchmark (comparison of the model with Facenet)

• Sviataslau Shchaurouski

  • Making client
  • Communication between client, model and index (including REST server for client).
  • Creating dockerfiles and docker-compose files. Deployment on DigitalOcean.

• Anton Grigoryev

  • Creating Index
    • Implementing HNSW algorithm
    • Making REST server