def init(self, location, frame):
"""
Initial tracker
"""
left, top, width, height = location
self._tracker.init(frame, location)
label_object(frame, left, top, left + width, top + height, self._name)
def init(self, location, frame):
"""
Initial tracker
"""
left, top, width, height = location
rect = dlib.rectangle(left, top, left + width, top + height)
self._tracker.start_track(frame, rect)
label_object(frame, left, top, left + width, top + height, self._name)
def update(self, frame):
"""
Update tracker
"""
ret_val, pos = self._tracker.update(frame)
if ret_val:
left = int(pos[0])
top = int(pos[1])
right = int(pos[0] + pos[2])
bottom = int(pos[1] + pos[3])
label_object(frame, left, top, right, bottom, self._name)
def update(self, frame):
"""
Update tracker
"""
confidence = self._tracker.update(frame)
pos = self._tracker.get_position()
if confidence > self.threshold:
left = int(pos.left())
top = int(pos.top())
right = int(pos.right())
bottom = int(pos.bottom())
label_object(frame, left, top, right, bottom, self._name)
def recognize_face(self, frame, tolerance=0.5, shrink=0.25):
"""
Recognize face
"""
face_locations, face_encodings = self.detect_face(frame, shrink)
for location, face_encoding in zip(face_locations, face_encodings):
left, top, right, bottom = [int(i / shrink) for i in location]
distances = np.linalg.norm(self.encodings - face_encoding, axis=1)
if min(distances) < tolerance:
name = self.names[distances.argmin()]
else:
name = "Unknown"
label_object(frame, left, top, right, bottom, name)
def recognize_face(self, frame, tolerance=1, shrink=1):
"""
Recognize face
"""
faces = self.detect_face(frame, shrink)
for location in faces:
left, top, width, height = [int(i / shrink) for i in location]
right = left + width
bottom = top + height
img = cv2.cvtColor(frame[top:bottom, left:right], cv2.COLOR_BGR2GRAY)
aligned_face = align_face(img)
if aligned_face is not None:
img = cv2.resize(aligned_face, (self.image_size, self.image_size))
label, confidence = self.face_recognizer.predict(aligned_face)
confidence = confidence / 100
name = self.face_recognizer.getLabelInfo(label)
if confidence > tolerance:
name = "Unknown"
label_object(frame, left, top, right, bottom, name)
def recognize_track_face(frame, tolerance, tracking, shrink=0.25):
"""
Recognize and track faces
"""
face_locations, face_encodings = detect_face(frame, shrink)
tracks = []
for location, face_encoding in zip(face_locations, face_encodings):
top, right, bottom, left = [int(i / shrink) for i in location]
distances = face_recognition.face_distance(ENCODINGS, face_encoding)
if min(distances) < tolerance:
name = NAMES[distances.argmin()]
else:
name = "Unknown"
if not tracking:
label_object(frame, left, top, right, bottom, name)
return None
if tracking == "dlib":
track = DlibCorrelationTracker(name, 5)
else:
track = OpencvTracker(name, tracking)
track.init((left, top, right - left, bottom - top), frame)
tracks.append(track)
return tracks