This repo contains the source and data of famousfaces, an open source variant of Microsofts CelebsLikeMe. You take a picture of yourself and it presents you the most similar celebrities of a given set. You can easily add/remove faces and integrate the used react webapp into your own.
Getting started
Use your own images
How does it work?
Limitations
https://docs.docker.com/engine/installation/
https://docs.docker.com/compose/install/
git clone https://github.com/siavash9000/famousfaces.git
cd famousfaces
docker-compose pull
docker-compose build
docker-compose up
The startup cann took several minutes, please be patient. The Tensorflow Service needs some time for loading its model. The application should then be available under http://localhost:3000
You want to use your own images? Just add as them as jpeg files to celebritydata/images. To create new embeddings run
docker-compose -f build-embeddings.yml up
Then rebuild and start the application:
docker-compose build
docker-compose up
The core of Famousfaces relies on facenet. Facenet is a neural network model which computes an embedding for a given face. An embedding of a face is a vector with a very usefull characteristic: The more similar the faces are to each other, the closer the vectors are to each other in terms of cosine distance. Famousfaces uses this property of facenet embeddings to find the most similar faces from a given set of images.
The tensorflow version is fixed to 1.2 . Feel free to update and create a pull request! Thanks in Advance for your help!