Strapi Face Recognition

This project uses PostgreSQL's cube type and face-api.js to find faces in the images uploaded to Strapi and create Face records for them.

This allows the Face content type to used in queries to return content that contains that face (e.g. a list of pages that contain images for a person)

Important

This project is a proof-of-concept and as such will only work with a PostgreSQL backend

Setting up DB

Strapi doesn't configure PostgreSQL databases as part of the project generation so this needs to be done yourself.

In PostgreSQL use the following (change mydb, myuser, mypass to your preferred values)

create database mydb;
create user myuser with encrypted password 'mypass';
grant all privileges on database mydb to myuser;
alter DATABASE mydb OWNER TO admin;

How it works

• When bootstrapping the face-recognition plugin the vectors table which stores the raw face descriptions is created
• A lifecycle listener is also set up that gets called when media is created (updated & deleted coming soon). This is used to detect and get the face descriptions for storage
• Once a face has been found and the descriptions generated, these are stored in the vectors table
• The vectors table has a link to the image ID

How it will work (when I get round to further development)

• A property - face_id is used to link the vector to a specific Face entry and that Face in turn is linked to the created media
• The media uploader has been extended to help with the tagging of Faces
• As more and more media is added the vectors table is used to suggest faces

Testing it

• Set up PostgreSQL database as above
• Git clone the project and install the deps
• Update the config/database.js values to ensure that can connect to DB
• Run yarn develop (or use npm) to run the server (this will create the vectors table)
• Once strapi is set up and you've created admin account the upload functionality should be ready (uploaded images will take longer if face detected)
• In order to use the faces/searches endpoint you will need to enable the public role for the searches endpoint, this is done under Settings > Roles > Public > Face-recognition and checking search
• you can then POST an image (use form-data and file as field name) to http://localhost:1337/api/faces/searches and it will return all matching faces