face-detection

Generating named bounding boxes on known/unknown faces

Requirements

This script requires Python 3.6+

Install dependencies via pip:

pip install -r requirements.txt

Virtualenv

Step 1: Creating new virtual environment

Best practice is to set up development environment using virtualenv package

You can install it via pip:

pip install virtualenv

python -m venv venv

or

virtualenv -p python venv

Step 2: Activate virtual environment

For Linux and Mac:

source venv/bin/activate

For Windows:

venv\Scripts\activate.bat

Deactivate the environment with

deactivate

(Additional) Use: python -h venv for help

For more details about how to set up your virtual environment, read the Docs

Usage

Let the to NN detect and store known faces data:

  python encode_faces.py -e ENCODINGS_PICKLE_PATH -c cnn
      
  Options:
      -d --dataset Path to folder with dataset
      (May contain subfolders, each named as a concrete person you want to recognize)
      
      -e --encodings Path to .pickle file where to serialize face encodings
      -m --method Learning model: use 'cnn' or 'hog'
       [
         'hog' : (Histogram of oriented gradients) -> faster, less accurate
         'cnn' : (Convolutional Neural Network) -> slower, more accurate
       ]

Run the real-time face recognition (Web-camera is required):

  python pi_face_recognition.py -e ENCODINGS_PICKLE_PATH -c CASCADE_PATH
      
  Options:
      -e --encodings Path to .pickle file with known face encodings
      -c --cascade Path to .xml cascade classifier file