def _detect_faces(self, img):
"""Detects faces on the image.
Args:
img -- image
Returns:
faces -- LIST of np.array([x,y,w,h]) face coordinates
"""
if not self._face_cascade.empty():
gray = img_to_gray(img)
faces = self._face_cascade.detectMultiScale(gray, \
scaleFactor=1.3, \
minNeighbors=4, \
minSize=(30, 30), \
flags=cv2.CASCADE_SCALE_IMAGE)
if len(faces) == 0:
return []
return faces
def substract_bg(self, frame):
"""Apply Background Substraction from frame.
Args:
frame -- current frame
Returns:
fgmask -- foreground mask
"""
# Get FGMASK with MOG2
gray = img_to_gray(frame)
fgmask = self._bs_mog2.apply(gray, learningRate=self.learningRate)
# Elliptical Kernel for morphology func
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,( \
self.kernel_size, self.kernel_size))
# Open (remove white points from the background)
fgmask = cv2.morphologyEx(fgmask, cv2.MORPH_OPEN, kernel)
# Close (remove black points from the object)
fgmask = cv2.morphologyEx(fgmask, cv2.MORPH_CLOSE, kernel)
# Dilation alg (increases white regions size)
fgmask = cv2.dilate(fgmask, kernel, iterations=self.iters)
return fgmask
def track_faces(self, img):
"""Apply oval mask over faces detected in the image.
Args:
img -- image
Updates:
self.faces -- list of faces detected
self.gc_rect -- coords. tuple for GrabCut algorithm (x, y, w, h)
Returns:
fgmask -- binary mask with faces highlighted with oval
"""
faces = self._detect_faces(img)
if faces != []: # Face coords detected
self.faces = faces
self.gc_rect = self._faces_to_gc_rect(faces)
else: # Default values
self.faces = self.def_face
self.gc_rect = self._faces_to_gc_rect(self.faces, 1)
fgmask = np.zeros((self.height, self.width, 3), np.uint8)
fgmask = draw_ellipse(fgmask, self.faces)
fgmask = img_to_gray(fgmask)
return fgmask
