def reload_images(self, group_method, img_list):
"""
Reloads the image list by replacing the comparative values with those
that the chosen grouping method expects.
:param group_method: str name of the grouping method that will be used.
:param img_list: image list that has been sorted by one of the sort
methods.
:return: img_list but with the comparative values that the chosen
grouping method expects.
"""
input_dir = self.args.input_dir
print("Preparing to group...")
if group_method == 'group_blur':
temp_list = [[x, self.estimate_blur(cv2.imread(x))] for x in tqdm(
self.find_images(input_dir), desc="Reloading", file=sys.stdout)
]
elif group_method == 'group_face':
temp_list = [[
x, face_recognition.face_encodings(cv2.imread(x))
] for x in tqdm(
self.find_images(input_dir), desc="Reloading", file=sys.stdout)
]
elif group_method == 'group_face_cnn':
temp_list = []
for x in tqdm(self.find_images(input_dir),
desc="Reloading",
file=sys.stdout):
d = face_alignment.Extract(
cv2.imread(x),
'dlib-cnn',
True,
input_is_predetected_face=True).landmarks
temp_list.append([
x,
np.array(d[0][1]) if len(d) > 0 else np.zeros((68, 2))
])
elif group_method == 'group_face_yaw':
temp_list = []
for x in tqdm(self.find_images(input_dir),
desc="Reloading",
file=sys.stdout):
d = face_alignment.Extract(
cv2.imread(x),
'dlib-cnn',
True,
input_is_predetected_face=True).landmarks
temp_list.append(
[x, self.calc_landmarks_face_yaw(np.array(d[0][1]))])
elif group_method == 'group_hist':
temp_list = [[
x,
cv2.calcHist([cv2.imread(x)], [0], None, [256], [0, 256])
] for x in tqdm(
self.find_images(input_dir), desc="Reloading", file=sys.stdout)
]
else:
raise ValueError("{} group_method not found.".format(group_method))
return self.splice_lists(img_list, temp_list)
def detect_faces(frame, detector, verbose, rotation=0, mtcnn_kwargs=None):
""" Detect faces and draw landmarks in an image """
face_detect = face_alignment.Extract(frame, detector, mtcnn_kwargs,
verbose)
for face in face_detect.landmarks:
ax_x, ax_y = face[0][0], face[0][1]
right, bottom = face[0][2], face[0][3]
landmarks = face[1]
yield DetectedFace(frame[ax_y:bottom, ax_x:right],
rotation,
ax_x,
right - ax_x,
ax_y,
bottom - ax_y,
landmarksXY=landmarks)
def sort_face_yaw(self):
input_dir = self.args.input_dir
img_list = []
for x in tqdm(self.find_images(input_dir),
desc="Loading",
file=sys.stdout):
d = face_alignment.Extract(
cv2.imread(x),
'dlib-cnn',
True,
input_is_predetected_face=True).landmarks
img_list.append(
[x, self.calc_landmarks_face_yaw(np.array(d[0][1]))])
print("Sorting by face-yaw...")
img_list = sorted(img_list, key=operator.itemgetter(1), reverse=True)
return img_list
def detect_faces(frame, detector, verbose, rotation=0, mtcnn_kwargs=None):
""" Detect faces and draw landmarks in an image """
face_detect = face_alignment.Extract(frame, detector, mtcnn_kwargs,
verbose)
# --- Changes ---
# If we have a center, take the face that is closest to previous center
if detect_faces.prev_center is None:
print('WARNING: Static variable in detect_faces, only extract one '
'sequence per faceswap.py call')
# Find biggest face by looking at y bounding box size, 0 is biggest
detect_landmarks = face_detect.landmarks
detect_landmarks = sorted(detect_landmarks,
key=lambda x: x[0][1] - x[0][3])
face_detect.landmarks = [detect_landmarks[0]]
else:
closest = float('inf')
closest_face_idx = -1
closest_center = None
for i, landmark in enumerate(face_detect.landmarks):
center = compute_center(landmark)
length = np.linalg.norm(
np.array(center) - np.array(detect_faces.prev_center))
if length < closest:
closest = length
closest_face_idx = i
closest_center = center
detect_faces.prev_center = closest_center
face_detect.landmarks = [face_detect.landmarks[closest_face_idx]]
detect_faces.prev_center = compute_center(face_detect.landmarks[0])
# --- Changes ---
for face in face_detect.landmarks:
ax_x, ax_y = face[0][0], face[0][1]
right, bottom = face[0][2], face[0][3]
landmarks = face[1]
yield DetectedFace(frame[ax_y:bottom, ax_x:right],
rotation,
ax_x,
right - ax_x,
ax_y,
bottom - ax_y,
landmarksXY=landmarks)
def sort_face_yaw(self):
""" Sort by yaw of face """
input_dir = self.args.input_dir
img_list = []
for img in tqdm(self.find_images(input_dir),
desc="Loading",
file=sys.stdout):
landmarks = face_alignment.Extract(
input_image_bgr=cv2.imread(img),
detector='dlib-cnn',
verbose=True,
input_is_predetected_face=True).landmarks
img_list.append(
[img, self.calc_landmarks_face_yaw(np.array(landmarks[0][1]))])
print("Sorting by face-yaw...")
img_list = sorted(img_list, key=operator.itemgetter(1), reverse=True)
return img_list
def sort_face_cnn_dissim(self):
""" Sort by dlib CNN dissimilarity """
input_dir = self.args.input_dir
print("Sorting by face-cnn dissimilarity...")
img_list = []
for img in tqdm(self.find_images(input_dir),
desc="Loading",
file=sys.stdout):
landmarks = face_alignment.Extract(
input_image_bgr=cv2.imread(img),
detector='dlib-cnn',
verbose=True,
input_is_predetected_face=True).landmarks
img_list.append([
img,
np.array(landmarks[0][1]) if landmarks else np.zeros((68, 2)),
0
])
img_list_len = len(img_list)
for i in tqdm(range(0, img_list_len - 1),
desc="Sorting",
file=sys.stdout):
score_total = 0
for j in range(i + 1, len(img_list)):
if i == j:
continue
fl1 = img_list[i][1]
fl2 = img_list[j][1]
score_total += np.sum(np.absolute((fl2 - fl1).flatten()))
img_list[i][2] = score_total
print("Sorting...")
img_list = sorted(img_list, key=operator.itemgetter(2), reverse=True)
return img_list
def sort_face_cnn(self):
""" Sort by dlib CNN similarity """
input_dir = self.args.input_dir
print("Sorting by face-cnn similarity...")
img_list = []
for img in tqdm(self.find_images(input_dir),
desc="Loading",
file=sys.stdout):
landmarks = face_alignment.Extract(
input_image_bgr=cv2.imread(img),
detector='dlib-cnn',
verbose=True,
input_is_predetected_face=True).landmarks
img_list.append([img, np.array(landmarks[0][1])
if landmarks
else np.zeros((68, 2))])
img_list_len = len(img_list)
for i in tqdm(range(0, img_list_len - 1),
desc="Sorting",
file=sys.stdout):
min_score = float("inf")
j_min_score = i + 1
for j in range(i + 1, len(img_list)):
fl1 = img_list[i][1]
fl2 = img_list[j][1]
score = np.sum(np.absolute((fl2 - fl1).flatten()))
if score < min_score:
min_score = score
j_min_score = j
(img_list[i + 1],
img_list[j_min_score]) = (img_list[j_min_score],
img_list[i + 1])
return img_list
def sort_face_cnn(self):
input_dir = self.args.input_dir
print("Sorting by face-cnn similarity...")
img_list = []
for x in tqdm(self.find_images(input_dir),
desc="Loading",
file=sys.stdout):
d = face_alignment.Extract(
cv2.imread(x),
'dlib-cnn',
True,
input_is_predetected_face=True).landmarks
img_list.append(
[x, np.array(d[0][1]) if len(d) > 0 else np.zeros((68, 2))])
img_list_len = len(img_list)
for i in tqdm(range(0, img_list_len - 1),
desc="Sorting",
file=sys.stdout):
min_score = float("inf")
j_min_score = i + 1
for j in range(i + 1, len(img_list)):
fl1 = img_list[i][1]
fl2 = img_list[j][1]
score = np.sum(np.absolute((fl2 - fl1).flatten()))
if score < min_score:
min_score = score
j_min_score = j
img_list[i + 1] = img_list[j_min_score]
img_list[j_min_score] = img_list[i + 1]
return img_list