def main():
# scaling up & down
src = np.array(Image.open(root / 'fig/castle.jpg'))
h, w, c = src.shape
padding = np.zeros((h, 4, c), dtype=np.uint8)
scale_down = seam_carving.resize(src, (w - 200, h))
scale_up = seam_carving.resize(src, (w + 200, h))
merged = np.hstack((src, padding, scale_down, padding, scale_up))
Image.fromarray(merged).show()
# forward energy vs backward energy
src = np.array(Image.open(root / 'fig/bench.jpg'))
h, w, c = src.shape
padding = np.zeros((h, 4, c), dtype=np.uint8)
backward = seam_carving.resize(src, (w - 200, h))
forward = seam_carving.resize(src, (w - 200, h), energy_mode='forward')
merged = np.hstack((src, padding, backward, padding, forward))
Image.fromarray(merged).show()
# object removal
src = np.array(Image.open(root / 'fig/beach.jpg'))
h, w, c = src.shape
mask = np.array(Image.open(root / 'fig/beach_girl.png').convert('L'))
dst = seam_carving.remove_object(src, mask)
padding = np.zeros((h, 4, c), dtype=np.uint8)
merged = np.hstack((src, padding, dst))
Image.fromarray(merged).show()
def main():
parser = argparse.ArgumentParser()
parser.add_argument('src', type=str)
parser.add_argument('-o', dest='dst', type=str, default='a.jpg')
parser.add_argument('--keep', type=str, default=None)
parser.add_argument('--drop', type=str, default=None)
parser.add_argument('--dw', type=int, default=0)
parser.add_argument('--dh', type=int, default=0)
parser.add_argument('--energy', type=str, default='backward',
choices=['backward', 'forward'])
parser.add_argument('--order', type=str, default='width-first',
choices=['width-first', 'height-first', 'optimal'])
args = parser.parse_args()
try:
print('Loading source image from {}'.format(args.src))
src = np.array(Image.open(args.src))
drop_mask = None
if args.drop is not None:
print('Loading drop_mask from {}'.format(args.drop))
drop_mask = np.array(Image.open(args.drop).convert('L'))
keep_mask = None
if args.keep is not None:
print('Loading keep_mask from {}'.format(args.keep))
keep_mask = np.array(Image.open(args.keep).convert('L'))
print('Performing seam carving...')
start = time.time()
if drop_mask is not None:
dst = seam_carving.remove_object(src, drop_mask, keep_mask)
else:
src_h, src_w, _ = src.shape
dst = seam_carving.resize(src, (src_w + args.dw, src_h + args.dh),
args.energy, args.order, keep_mask)
print('Done at {:.4f} second(s)'.format(time.time() - start))
print('Saving output image to {}'.format(args.dst))
Path(args.dst).parent.mkdir(parents=True, exist_ok=True)
Image.fromarray(dst).save(args.dst)
except Exception as e:
print(e)
exit(1)
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--src', type=str, default=None)
parser.add_argument('--mask', type=str, default=None)
parser.add_argument('--dst', type=str, default=None)
parser.add_argument('--delta-col', type=int, default=0)
args = parser.parse_args()
src = cv2.imread(args.src)
if args.mask:
mask = cv2.imread(args.mask, cv2.IMREAD_GRAYSCALE)
dst = seam_carving.remove_object(src, mask)
else:
if args.delta_col > 0:
dst = seam_carving.insert_seams(src, args.delta_col)
else:
dst = seam_carving.remove_seams(src, abs(args.delta_col))
cv2.imwrite(args.dst, dst)
mask = io.imread('imgs/wyeth_mask.jpg', as_grey=True)
mask = util.img_as_bool(mask)
plt.subplot(1, 2, 1)
plt.title('Original Image')
plt.imshow(image)
plt.subplot(1, 2, 2)
plt.title('Mask of the object to remove')
plt.imshow(mask)
plt.show()
from seam_carving import remove_object
# Use your function to remove the object
out = remove_object(image, mask)
plt.subplot(2, 2, 1)
plt.title('Original Image')
plt.imshow(image)
plt.subplot(2, 2, 2)
plt.title('Mask of the object to remove')
plt.imshow(mask)
plt.subplot(2, 2, 3)
plt.title('Image with object removed')
plt.imshow(out)
plt.show()
def detect_facial_feature(self, img, visible=False):
face_cascade = cv2.CascadeClassifier('cascades/frontalFace.xml')
eye_cascade = cv2.CascadeClassifier('cascades/Eyes_cascade.xml')
nose_cascade = cv2.CascadeClassifier('cascades/Nose_cascade.xml')
mouth_cascade = cv2.CascadeClassifier('cascades/Mouth_cascade.xml')
roi_data = {}
overlay = img.copy()
gray = cv2.cvtColor(overlay, cv2.COLOR_BGR2GRAY)
faces = self.__get_best_feature_roi(gray,
face_cascade,
k_mean=5,
scale=1.3)
for (x, y, w, h) in faces:
face = img[y:y + h, x:x + w]
roi_data['face'] = face
roi_gray = gray[y:y + h, x:x + w]
roi_color = overlay[y:y + h, x:x + w]
face_img = roi_color.copy()
cv2.rectangle(overlay, (x, y), (x + w, y + h), (255, 0, 0), 2)
face_center = self.__get_center(left_top_pixel=(x, y),
right_bottom_pixel=(x + w, y + h))
cv2.circle(overlay, face_center, 2, (255, 0, 0), -1)
print('Face: ', face_center)
eyes = self.__get_best_feature_roi(roi_gray,
eye_cascade,
k_mean=3,
scale=1.3)
for (ex, ey, ew, eh) in eyes:
roi_data['eyes'] = face[ey:ey + eh, ex:ex + ew]
cv2.rectangle(roi_color, (ex, ey), (ex + ew, ey + eh),
(0, 255, 0), 2)
#cv2.circle(roi_color, (int((ey+eh)/2), int((ex+ew)/2)), 2, (0, 255, 0), -1)
eyes_center = self.__get_center(left_top_pixel=(ex, ey),
right_bottom_pixel=(ex + ew,
ey + eh))
cv2.circle(roi_color, eyes_center, 2, (0, 255, 0), -1)
print('Eyes: ', eyes_center)
nose = self.__get_best_feature_roi(roi_gray,
nose_cascade,
k_mean=3,
scale=1.3)
for (nx, ny, nw, nh) in nose:
roi_data['nose'] = face[ny:ny + nh, nx:nx + nw]
cv2.rectangle(roi_color, (nx, ny), (nx + nw, ny + nh),
(0, 0, 255), 2)
nose_center = self.__get_center(left_top_pixel=(nx, ny),
right_bottom_pixel=(nx + nw,
ny + nh))
cv2.circle(roi_color, nose_center, 2, (0, 0, 255), -1)
print('Nose: ', nose_center)
mouth = self.__get_best_feature_roi(roi_gray,
mouth_cascade,
k_mean=50,
k_mean_step=10,
scale=1.3)
for (mx, my, mw, mh) in mouth:
roi_data['mouth'] = face[my:my + mh, mx:mx + mw]
cv2.rectangle(roi_color, (mx, my), (mx + mw, my + mh),
(255, 255, 0), 2)
mouth_center = self.__get_center(left_top_pixel=(mx, my),
right_bottom_pixel=(mx + mw,
my + mh))
cv2.circle(roi_color, mouth_center, 2, (255, 255, 0), -1)
print('Mouth: ', mouth_center)
if len(eyes) == 1 and len(nose) == 1:
ex, ey, ew, eh = eyes[0]
nx, ny, nw, nh = nose[0]
roi_data['right_cheek'] = face[ey + eh:int((ny + ny + nh) / 2),
ex:nx]
roi_data['left_cheek'] = face[ey + eh:int((ny + ny + nh) / 2),
nx + nw:ex + ew]
cv2.rectangle(roi_color, (ex, ey + eh),
(nx, int((ny + ny + nh) / 2)), (255, 0, 255), 2)
cv2.rectangle(roi_color, (nx + nw, ey + eh),
(ex + ew, int(
(ny + ny + nh) / 2)), (255, 0, 255), 2)
removed_features_on_face = face
# if len(eyes) == 1:
# removed_features_on_face = self.feature_smoothing(removed_features_on_face, eyes[0])
# if len(nose) == 1:
# removed_features_on_face = self.feature_smoothing(removed_features_on_face, nose[0])
# if len(mouth) == 1:
# removed_features_on_face = self.feature_smoothing(removed_features_on_face, mouth[0])
# roi_data['skin'] = removed_features_on_face
if len(eyes) == 1:
pass
#removed_features_on_face = seam_carving.remove_object(removed_features_on_face, eyes[0])
if len(nose) == 1:
removed_features_on_face = seam_carving.remove_object(
removed_features_on_face, nose[0])
if len(mouth) == 1:
removed_features_on_face = seam_carving.remove_object(
removed_features_on_face, mouth[0])
roi_data['skin'] = removed_features_on_face
if visible:
h, w = overlay.shape[:2]
if h > 1200 or w > 1200:
img = cv2.resize(overlay,
dsize=None,
fx=0.4,
fy=0.4,
interpolation=cv2.INTER_AREA)
cv2.imshow('Facial Features', overlay)
#cv2.imshow('Mask', self.__filter_skin(face_img))
cv2.waitKey()
cv2.destroyAllWindows()
return roi_data
def main():
parser = argparse.ArgumentParser()
parser.add_argument('src', type=str)
parser.add_argument('-o', dest='dst', type=str, required=True)
parser.add_argument('--keep', type=str, default=None)
parser.add_argument('--drop', type=str, default=None)
parser.add_argument('--dw', type=int, default=0)
parser.add_argument('--dh', type=int, default=0)
parser.add_argument('--energy',
type=str,
default='backward',
choices=['backward', 'forward'])
parser.add_argument('--order',
type=str,
default='width-first',
choices=['width-first', 'height-first', 'optimal'])
parser.add_argument('--face', action='store_true')
args = parser.parse_args()
try:
print('Loading source image from {}'.format(args.src))
src = np.array(Image.open(args.src))
drop_mask = None
if args.drop is not None:
print('Loading drop_mask from {}'.format(args.drop))
drop_mask = np.array(Image.open(args.drop).convert('L'))
keep_mask = None
if args.keep is not None:
print('Loading keep_mask from {}'.format(args.keep))
keep_mask = np.array(Image.open(args.keep).convert('L'))
if args.face:
# Face detection using face++ API
with open(Path(__file__).parent / 'api_config.json') as f:
api_config = json.load(f)
with open(args.src, 'rb') as f:
img_data = f.read()
image_base64 = base64.b64encode(img_data)
url = 'https://api-us.faceplusplus.com/facepp/v3/detect'
data = {
'api_key': api_config['api_key'],
'api_secret': api_config['api_secret'],
'image_base64': image_base64
}
response = requests.post(url, data)
data = response.json()
src_h, src_w, _ = src.shape
if keep_mask is None:
keep_mask = np.zeros((src_h, src_w), dtype=np.bool)
for face in data['faces']:
rect = face['face_rectangle']
x1 = rect['left']
y1 = rect['top']
w = rect['width']
h = rect['height']
keep_mask[y1:y1 + h, x1:x1 + w] = True
print('Performing seam carving...')
start = time.time()
if drop_mask is not None:
dst = seam_carving.remove_object(src, drop_mask, keep_mask)
else:
src_h, src_w, _ = src.shape
dst = seam_carving.resize(src, (src_w + args.dw, src_h + args.dh),
args.energy, args.order, keep_mask)
print('Done at {:.4f} second(s)'.format(time.time() - start))
print('Saving output image to {}'.format(args.dst))
Path(args.dst).parent.mkdir(parents=True, exist_ok=True)
Image.fromarray(dst).save(args.dst)
except Exception as e:
print(e)
exit(1)