/facial_transformation

This repository will provide a way of transforming an already detected face and output it in a predefined template.

Primary LanguageC++

Face Aligner

This repository will provide a way of transforming an already detected face and output it in a predefined template.

Class

Constructor

The constructor for the FaceAligner class needs three arguments:

  1. String: Path for the landmark model (See FaceLandmarkDetector class).
  2. Int: Size of the template image where the aligned face will be shown.
  3. Float: Value from 0 to 1 giving the x and y coordinate of the left eye in the template image.

Functions

  1. Detect: Detects the landmarks for the given image:

    1. cv::Mat: Image to detect
    2. cv::Rect: Rectangle that bounds the face
    3. dlib::full_object_detection: Struct where the detected landmark points will be saved.

  2. Align: Aligns the given image according to the points given and outputs it to the given image:

    1. dlib::full_object_detection: Detected points used to align image
    2. cv::Mat: Image to align
    3. cv::Mat: Image to output the aligned face

  3. DetectAndAlign: Detects the landmarks for the given image and face, aligns the image and outputs it to the given Mat

    1. cv::Mat: Image to detect and align
    2. cv::Rect: Rectangle that bounds the face
    3. cv::Mat: Image to output the aligned face

FaceLandmarkDetector

This class will detect the landmarks for the given face and the given image.

Constructor

The constructor for this class needs the path for the landmark detector model to use. There are two in this repository:

  1. shape_predictor_5_face_landmarks.dat
  2. shape_predictor_68_face_landmarks.dat

Both of these were downloaded from this repository. We recomend the use of the shape_predictor_5_face_landmarks.

Functions

  1. GetFaceLandmark: Detects the important points of a given image.
    1. cv::Mat: Image to detect
    2. cv::Rect: Rectangle that bounds the face
    3. dlib::full_object_detection: Struct where the detected landmark points will be saved.

Aligner

This class will align an image based on given points.

Constructor

The constructor of this class needs two arguments:

  1. Int: Size of the template image where the aligned face will be shown.
  2. Float: Value from 0 to 1 giving the x and y coordinate of the left eye in the template image.

Functions

  1. AlignImage: Aligns a given image based on the points given and outputs it to the given image.
    1. dlib::full_object_detection: Detected points used to align image
    2. cv::Mat: Image to align
    3. cv::Mat: Image to output the aligned face