Learns and identifies faces, with OpenCV and Python. Fully customizable and extendable.

The goal is to have an extendable face recognition application, that is easy to use and to customize. The algorithm should learn continuously known faces and save additional training or validation sets for retraining the model and enhancing the accuracy.

1. Clone the git repository with the submodules. If you're working with the Terminal, here's the convenient code for you:

git clone --recursive https://github.com/n1g1r1/cv-face-recognition

2. Switch into cloned project folder and install necessary dependencies:

cd cv-face-recognition && pip install -r requirements.txt

3. Run the python script:

python main.py

.. or if you're that kind of person, here's a convenient one-liner:

git clone --recursive https://github.com/n1g1r1/cv-face-recognition && cd cv-face-recognition && pip install -r requirements.txt && python main.py

The application comes with a handy default config.yml file with default values. Check this file out and customize it if you want. You can also make multiple config files for your purposes and change the config file name in main.py.

Call the script with

python main.py

You can customize the settings of the application by editing the config.yml file.

# General settings
general:
    face_classifier:    'haar' # Choose: lbp, haar
    recognizer:         'lbp' # Choose: lbp, eigen, fisher
    identifier_method:  'local' # Choose: local, remote
    capture_faces:      True
    train_model:        False
    classify:           True
    resize_factor:      0.5

# Training
training:
    set_path:         'data/training'
    number_of_images: 20
    resize:           False
    model_path:       'model'
    model_filename:   'LBPHData.xml'

# Validation
validation:
    make_set:         False
    set_path:         'data/validation'
    number_of_images: 20

# Test set
test:
    make_set:         True
    set_path:         'data/test'

# Server configuration for classification
server:
    url:              '' # Trailing slash
    classify:         ''
    train:            ''
    update:           ''

To detect faces in general, you can choose between two classifiers at face_classifier:

• lbp Local Binarization Pattern Histogram
• haar Haar Cascades

To train and recognize the algorithm locally, you can choose between those three options at recognizer:

• lbp Local Binarization Pattern Histogram
• eigen Eigenfaces
• fisher Fisherfaces

You can even have your own remote recognizer that returns a JSON formatted label and confidence value. If so, fill in an url to the server, as well as the paths to the classify, train and update API interface in the server section. For example, you might have the server address http://recognize.faceidentifier.io, so you might fill in the section like the following:

# Server configuration for classification
server:
    url:              'http://recognize.faceidentifier.io/' # Trailing slash
    classify:         'classify'
    train:            'model/train'
    update:           'model/update'

You might want to also customize the classifier_online module to even match more your needs.

More information:

• online classifier module: https://github.com/n1g1r1/cv-module-face-classifier-online
• @quving's face classifier server API: https://github.com/Quving/cv-face-classifier