Face Trigger - Process package

Contains code for performing face detection, landmark detection and face alignment.

Submodules

face_trigger.process.post_process module

Module for performing Face detection, Face Alignment and Landmark Detection

class face_trigger.process.post_process.FaceAlign(left_eye_offset=(0.35, 0.35), final_width=256, final_height=None)

Align a face by performing affine transformations. Inspired from ‘Face Alignment with OpenCV and Python’ https://www.pyimagesearch.com/2017/05/22/face-alignment-with-opencv-and-python/

align(img, landmarks)

Align the given image according to given face landmarks - 1. Eyes are aligned on a horizontal axis 2. Face is scaled to keep eye centers on the same offset location w.r.t the face 3. Face is centered on the center of mass of the eyes.

Parameters:img (numpy.ndarray) – rgb or gray-scale image/frame
Landmarks:the x,y-coordinates of the facial landmarks
Returns:aligned image of final dimensions
Return type:numpy.ndarray
class face_trigger.process.post_process.FaceDetector(face_area_threshold=0.25)

HOG-based frontal face detector class.

detect(gray_frame)

Detect faces in the frame. If multiple faces are detected, only the largest face is taken into consideration.

Parameters:gray_frame (numpy.ndarray) – grayscale image that might include a face
Returns:bounding box coordinates signifying the location of the face
Return type:dlib.rectangle
detect_unbounded(gray_frame)

Detect all faces in the frame

Parameters:gray_frame (numpy.ndarray) – grayscale image that might faces
Returns:bounding box coordinates signifying the location of the faces
Return type:list of dlib.rectangle
class face_trigger.process.post_process.LandmarkDetector(predictor_path=None)

A landmark detector that annotates face bounding boxes with 5 landmarks

predict(bounding_box, grayImg)

Provides an array of tuples for facial landmarks, predicted within a bounding box

Parameters:
  • bounding_box (dlib.rectangle) – bounding box coordinates in dlib format
  • grayImg (numpy.ndarray) – grayscale image
Returns:

5-tuple
Return type:

dlib.full_object_detection

Module contents