Note: NOT ACTIVE

Face Recognizer

Face Recognizer identifies the face of an individual by their name with the help of their facial features.

Face Recognizer uses deep learning algorithms to compare a live capture or digital image with the stored faceprints(also known as datasets) to verify an identity.

The algorithm used for classification is k-NN model i.e. k-Nearest Neighbor classifier. It uses Euclidean distance to compare images for similarity.

Prerequisites

Install cmake library

pip install cmake

Clone the dlib repository

git clone https://github.com/davisking/dlib.git

Make build directory

cd dlib
mkdir build
cd build

Install setup.py

cmake ..
cmake --build
cd ..
python setup.py install

Setup

Windows:

pip install -r requirements.txt

Linux:

• Setup environment

  sudo apt-get update
  sudo apt-get upgrade
  

• Install dependencies for OpenCV

  sudo apt-get install build-essential cmake unzip pkg-config
  sudo apt-get install libjpeg-dev libpng-dev libtiff-dev
  sudo apt-get install libjasper-dev
  sudo apt-get install libavcodec-dev libavformat-dev libswscale-dev libv4l-dev
  sudo apt-get install libxvidcore-dev libx264-dev
  sudo apt-get install libgtk-3-dev
  sudo apt-get install libatlas-base-dev gfortran
  sudo apt-get install python3.6-dev
  sudo apt-get install libboost-all-dev
  

• Install OpenCV

  • OpenCV

    pip install opencv-python
    

  • OpenCV Contrib

    pip install opencv-contrib-python
    

• Install other project dependencies pip install -r requirements.txt

  Further error can be solved

  • On installing opencv: https://www.pyimagesearch.com/2018/05/28/ubuntu-18-04-how-to-install-opencv/
  • On dlib: https://www.pyimagesearch.com/2017/03/27/how-to-install-dlib/

Set the path of the input images in the code and run the following command:

python FaceRecognizer.py

How Does It Work?

* An image that contains the face to be recognized is loaded (e.g., input image).

* The face is detected and segmented.

* The face is aligned if it is not aligned straight

* Landmarks (e.g., 68 (x, y) coordinates) are detected from the facial structure.

* The detected face landmarks are encoded (e.g., 128-d embeddings of the image are created).

* The input image[encoding] is passed to the k-NN model for classification.

* The k-NN model returns the name with the highest precision

Outputs:

• Input Image - The image in which faces are to be recognized:

• Original Image - The cropped input image

!

• Aligned Image - Aligning the image increases the efficiency of the algorithm

• Landmarks - Shows the landmarks of the detected faces

• Detected Face - The face is recognized and the name of the recognized face is displayed along with the face. If the face does not belong to the dataset then the face is labeled as Unknown.

Another Recognition:

• Input Image:

• Original Image:

• Aligned Image:

• Landmarks :

• Detected Face :

encode-faces.py

It is used to create 128-d face embeddings of the input image as well as custom dataset. These embeddings are used to compare input image(embeddings) with the dataset(embeddings). The one with the highest votes is preferred.

DAT file

Link to Download Dat File

Resources

• I got big help from pyimagesearch. It cleared most of my concepts with very Ease and practical examples.
• Adam Geitgey : How Face Recognition works
• Face Recognition
• Custom Dataset: helped me to create my own dataset.
• Must Read Post
• Deep Face Recognition
• Facial Landmarks
• Facial Alignment
• Face Recognition for Beginners
• Face Recognition Basics