Face Recognition Pipeline

This repository hosts a complete face recognition pipeline using dlib pre-trained models.

Installation

Create a conda or virtualenv environment with a Python 3.7 base, and install the dependencies named in requirements.txt

conda create -n facerecog python=3.7
conda activate facerecog
pip install -r requirements.txt

How to use

1. Create a face databank

You need to create a face databank by putting in a directory the following structure:

person_1/
    img1.jpg
    img2.jpg
    ...
person_2/
    image1.png
    image2.png
    ...
...

The amount of images per person does not need to be the same. The name of the image within the folder also doesnt' matter, as long as it is a png, jpg or gif file.

2. Enroll people into the system

To enroll all the persons in your face databank you need to call the following:

$ python enroll.py --dataset <path to dataset>

If an image has multiple faces, the enrolling script will consider the person's face the one who is more horizontally aligned to the center of the image.

3. Recognize people

Finally to recognize people within an image you can call:

$ python recognize.py -i <path to image>

Or to recognize people from a file with the list of image paths you may call:

$ python recognize_list.py -i <path to image list>

The results will be stored in a numpy file which contains a dictionary with the results per image.