Face Detection

This is an HTTP service that provides functions to detect faces in images. It includes:

• Face Detection Resource

For understanding the APIs and resources please refer to the folder resources or the explanation here

Folders

• bin: executable file creating the necessary folders and copying models before building the docker
• mtcnn: Tensorflow model for detecting faces
• resources: resources for different APIs

Docker

In this part we explain how to build and run the docker.

Build a Docker container

sudo bin/docker_build_context.sh
sudo docker build --tag=face_detection:0.0.1 build/docker

Run Docker container

Just CPU:

docker run -d --name face_detection -p 8000:80

After starting the container the service should listen on 127.0.0.1 port 8000.

The number of Gunicorn workers can be configured by setting the NUM_WORKER environment variable when running the container, e.g. -e NUM_WORKER=2.

Start service manually

For debugging it can be helpful to start the service manually. Run the container but overwrite the entrypoint with a Bash shell (you need to modify the version manually instead of 0.0.1):

docker run -it -p 8000:80 --entrypoint=/bin/bash face_detection:0.0.1

This starts the container and opens a shell but does not start the service. Start the service manually:

cd /app
gunicorn --workers 1 --worker-class gevent --bind 0.0.0.0:80 main:app

Start service without Docker

FACE_DETECTION_CONFIG=service-dev.yaml gunicorn --workers 1 --worker-class gevent --bind 0.0.0.0:8000 main:app

Login to the Docker container

Lookup container ID:

    docker container ps
    CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS                  NAMES
    7932a3814453        friendlyhello       "python app.py"     16 seconds ago      Up 15 seconds       0.0.0.0:4000->80/tcp   musing_robinson

Open a shell on the container:

docker exec -it c1de50a17e8d /bin/bash

Cleanup

Remove all containers and images:

docker rm $(docker ps -a -q)
docker rmi $(docker images -q)

Start service locally

The service requires a YAML config file. The config file path can be specified by setting an environment variable FACE_DETECTION_SERVICE_CONFIG. The default is to load service.yaml from the current directory.

Start service locally:

PYTHONUNBUFFERED=TRUE FACE_DETECTION_CONFIG=./service.yaml gunicorn --reload --workers 1 --worker-class gevent --bind 0.0.0.0:8080 main:app

Releasing a version

Build and push container to registry

git tag -a v#.##.#
./package-docker --push

Gunicorn

The Gunicorn configuration is described in Gunicorn settings.

The most important Gunicorn configuration parameters are:

• --reload - Restart workers when code changes. This should only be used during development
• --workers - The number of worker processes for handling requests
• --worker-class - The type of workers to use
• --bind - The socket and port to bind
• --access-logfile - Path of the access log file
• --error-logfile - Path of the error log file
• --daemon - Daemonize the Gunicorn process