Overview

Face Detection models:

1. SSD-Mobilenet
2. MTCNN
3. SSD-Inception

Face Embedding models:

1. Inception Resnet v1
2. Inception Resnet v2

Run the code in virtual environment (often ran into poor GPU utilization)

$ pip3 install -r requirements.txt

To run the Face Recognition test

$ python test.py

To run the Face Recognition test on multicam or multiple videos

$ python multicam.py

To benchmark the current Face Recognition architecture

$ python evaluate.py --path path/to/folder

To register new person

There are 2 ways:

1. Register from webcam: $ python register.py --source webcam --path 0

2. Register from folder: $ python register.py --source folder --path path/to/folder

The folder structure should be:

.
├── person1
│   ├── img1.jpg
│   ├── img2.jpg
│	└── ...
├── person2
│	├── img1.jpg
│   ├── img2.jpg
│	└── ...
└── ...

To tune model parameters

You can find the parameters in the config.py file.

APIs

To run the Face Registration API

$ python registration_api.py

To run the Face Recognition API

$ python recognition_api.py

Note: This code is compatible with Python >= 3.6.0. Although it might work with other Python versions with some minor changes