Face Recognition API

Tools

• Face Recognition model: deepface

• API: fastapi

Directory structure

This is how the project directory is structured:

├── app
│   ├── config.py
│   ├── main.py
│   └── utils.py
│   ├── data    # Database - Folder of images
│   │   ├── *.[jpg | jpeg]
|
├── query   # Query - Folder of images need to find id
│   ├── *.[jpg | jpeg]
|
├── docker-compose.yml
├── docker_build.sh
├── docker_run_it_test.sh
├── docker_run_server.sh
├── requirements.txt
└── uvicorn_run.sh

• data: "database" which contains all users image uniquely by their id/name.

• query: "query" which contains all images need to find identity base on the simmilarity between it with the image in database

• I use docker volume to manage persistent data instead of using disk.

Model configuration

• Face recognition model includes 2 modules: face detection and face recognition.

• For editting detector and recognizer configuration, see config.py for your specific configuration. Otherwise leaves as default.

Run

1. Install dependencies

pip install -r requirements.txt

2. Create ./app/data and ./query as directory structure above.

3. Run the server

uvicorn app.main:app --host 0.0.0.0 --port 80

• Add --reload flag to enable live mode.
• Go to localhost:{port}/docs for Swagger UI, check document for more information.

Docker

Directory structure in docker container:

├── /app
│   ├── app
│   │   ├── config.py
│   │   ├── main.py
│   │   ├── utils.py
|   |
│   ├── data  # Database - Folder of images
│   │   ├── *.[jpg | jpeg]
|   |
│   ├── query  # Query - Folder of images need to find id
│   │   ├── *.[jpg | jpeg]
|   |

1. Run the service

docker compose up --build -d

2. Run manually

Build the image:

bash docker_build.sh

Run the container

bash docker_run_server.sh