Face API

Face API with Docker, Tensorflow and PyTorch.

Models:

• Face Detection - TensorFlow (SSD-Mobilenet)
• Facial Emotion Recognition - PyTorch (Resnet50)
• Age and Gender Estimation - Keras (WideResnet)
• Facial Recognition - PyTorch (VGGFace2)

Requirements:

• Docker with Nvidia Support Nvidia-Docker
• CUDA 9 Guide

How to use:

In order to start the API the user must run the docker containers using the docker-compose command.

To build the docker images use:

docker-compose -f dockers/docker-compose.yml build

To run the docker containers use:

docker-compose -f dockers/docker-compose.yml up

The user can access the API using the following URL:

http://<IP>:<PORT>/predict/

It can be any IP (public or private) connected to the server, and each model is hosted in a different port e.g. 7001, 7002, 7003.

Test:

Change the host variable inside the test_api.py script with the IP in which the dockers are running and then run the following:

python test_api.py

Run the demo:

This demo calls the API to make predictions using the webcam.

Make sure the docker containers are running and change the host variable with the desired IP.

Run the demo using:

python demo.py