def original_image_processing(filename):
input_img = plt.imread(filename)
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade.
detector = Cascade(trained_file)
# Detect faces with scale factor to 1.2 and step ratio to 1
detected = detector.detect_multi_scale(img=input_img,
scale_factor=1.2,
step_ratio=1,
min_size=(50, 50),
max_size=(200, 200))
original_img = np.copy(input_img)
original_img = input_img / 255
i = 0
for d in detected:
i += 1
# Obtain the face cropped from detected coordinates
face = get_face_coordinates(d, input_img)
print("Face -", i)
plt.imshow(face)
plt.show()
blurred_face = gaussian(face, multichannel=True, sigma=8)
print("Blurred Face -", i)
plt.imshow(blurred_face)
plt.show()
resulting_image = merge_with_original(d, original_img, blurred_face)
return resulting_image
def get_face(img1, **kwargs):
"""
Given an image, and (optionally) a feature set, return a roi that is cropped into containing a face
:param img1:
:param kwargs:
:return: a cropped image of a face
"""
# create a detector from Cascade
if 'detector' in kwargs:
detector = kwargs['detector']
else:
if 'features_xml_url' in kwargs:
detector = Cascade(kwargs['features_xml_url'])
else:
detector = Cascade(
os.path.join(WORK_ROOT, 'RES', 'FD',
'lbpcascade_frontalface.xml'))
# detecting face in an image
detected = detector.detect_multi_scale(img=img1,
scale_factor=1.2,
step_ratio=1,
min_size=(60, 60),
max_size=(123, 123))
if len(detected) == 0:
return None
# print(f'img is {type(img)}, size is {img.shape}, {type(img.reshape((200, 180)))}, {len(detected)}')
c = detected[0]['c']
r = detected[0]['r']
return img1.reshape((200, 180))[c:c + detected[0]['height'],
r:r + detected[0]['width']]
def re():
dir_path = os.path.dirname(os.path.realpath(__file__))
cwd = os.getcwd()
print(cwd)
os.chdir(dir_path)
image1 = ""
image = io.imread("somefilename.jpg")
# Load the trained file from data
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade
detector = Cascade(trained_file)
detected = detector.detect_multi_scale(img=image,
scale_factor=1.2,
step_ratio=1,
min_size=(50, 50),
max_size=(100, 100))
resulting_image = image
for detected_face in detected:
face = get_face(image, detected_face)
blurred_face = gaussian(face, multichannel=True, sigma=10)
#show_image(face)
resulting_image = merge_blurry_face(image, detected_face, blurred_face)
show_image(resulting_image, "Blurred faces")
os.remove("somefilename.jpg")
return 'Its Done!'
def initialize_model(self, weights: Any=None) -> None:
if weights is None: # passing a non existent directory
weights = self.load_weight(weight_path=Path("fdagkjhkjhlkfytfkjhfg"))
else:
weights = self.load_weight(weight_path=Path(weights))
self.model = Cascade(weights)
class SkimageCascadeModel(BaseModel):
def __init__(self):
super(SkimageCascadeModel, self).__init__()
def load_weight(self, weight_path: Path) -> Any:
if weight_path.exists():
# TODO: Implement loading weight file from local path
pass
else:
logger.warning("No existing path was provided, using default weights")
weight_path = skimage_data.lbp_frontal_face_cascade_filename()
return weight_path
def initialize_model(self, weights: Any=None) -> None:
if weights is None: # passing a non existent directory
weights = self.load_weight(weight_path=Path("fdagkjhkjhlkfytfkjhfg"))
else:
weights = self.load_weight(weight_path=Path(weights))
self.model = Cascade(weights)
def predict(self, x) -> Iterable[Dict[str, int]]:
w, h = x.shape[:-1]
min_w, min_h = w * 0.01, h * 0.01
max_w, max_h = w * 0.9, h * 0.9
# detected contain {'c':int, 'r':int, 'width':int, 'height':int}
detected = self.model.detect_multi_scale(img=x,
scale_factor=1.2,
step_ratio=1,
min_size=(min_w, min_h),
max_size=(max_w, max_h))
return detected
def test_detector_astronaut():
# Load the trained file from the module root.
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade.
detector = Cascade(trained_file)
img = data.astronaut()
detected = detector.detect_multi_scale(img=img,
scale_factor=1.2,
step_ratio=1,
min_size=(60, 60),
max_size=(123, 123))
assert len(detected) == 1, 'One face should be detected.'
def test_detector_astronaut():
# Load the trained file from the module root.
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade.
detector = Cascade(trained_file)
img = data.astronaut()
detected = detector.detect_multi_scale(img=img,
scale_factor=1.2,
step_ratio=1,
min_size=(60, 60),
max_size=(123, 123))
assert len(detected) == 1, 'One face should be detected.'
def face_detections(image,
scale_factor=1.1,
step_ratio=1.3,
min_size=(10, 10),
max_size=(400, 400)):
# Load the trained file from data
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade
detector = Cascade(trained_file)
# Detect faces with min and max size of searching window
detected_faces = detector.detect_multi_scale(
img=image,
scale_factor=scale_factor,
step_ratio=step_ratio,
min_size=min_size,
max_size=max_size,
)
return detected_faces
def fd():
from skimage.feature import Cascade
from skimage import data, color
import matplotib.pyplot as plt
from matplotlib.patches import Rectangle
path = ''
img = plt.imread(path)
plt.axis('off')
plt.imshow(img)
train_set = data.lbp_frontal_face_cascade_filename()
detector = Cascade(train_set)
detected = detector.detect_multi_scale(img=img,
scale_factor=1.2,
step_ratio=1,
min_size=(10, 10),
max_size=(200, 200))
print('Detected')
def show_detected_face(result, detected, title="Detected Faces"):
plt.imshow(result)
img_desc = plt.gca()
plt.set_cmap('gray')
plt.title(title)
plt.axis('off')
for rec in detected:
img_desc.add_patch(
Rectangle((rec['c'], rec['r']),
rec['width'],
rec['height'],
fill=False,
color='g',
line_width=3))
plt.show()
show_detected_face(img, detected)
you want to have a really precise detector, you will have to train it yourself
using `OpenCV train cascade utility
<https://docs.opencv.org/2.4/doc/user_guide/ug_traincascade.html>`_.
"""
from skimage import data
from skimage.feature import Cascade
import matplotlib.pyplot as plt
from matplotlib import patches
# Load the trained file from the module root.
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade.
detector = Cascade(trained_file)
img = data.astronaut()
detected = detector.detect_multi_scale(img=img,
scale_factor=1.2,
step_ratio=1,
min_size=(60, 60),
max_size=(123, 123))
plt.imshow(img)
img_desc = plt.gca()
plt.set_cmap('gray')
for patch in detected:
#header files required
from skimage.feature import Cascade
from skimage import data
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle
#Getting the image
path = ''
img = plt.imread(path)
plt.axis('off')
plt.imshow(img)
#reading the dataset for facial detection
train_set = data.lbp_frontal_face_cascade_filename()
detector = Cascade(train_set)
detected = detector.detect_multi_scale(img=img,
scale_factor=1.2,
step_ratio=1,
min_size=(10, 10),
max_size=(200, 200))
#function to draw a rectangle around the detected faces
def show_detected_face(result, detected, title='Detected Faces'):
plt.imshow(result)
desc = plt.gca()
plt.set_cmap('gray')
plt.title(title)
plt.axis('off')
for rec in detected:
desc.add_patch(
def showDetectedFaces(image,detected):
plt.imshow(image)
img_desc = plt.gca()
plt.set_cmap('rainbow')
plt.axis('off')
for patch in detected:
img_desc.add_patch(patches.Rectangle((patch['c'], patch['r']), patch['width'], patch['height'], fill=False, color='r', linewidth=2))
plt.show()
#load the trained file from the module root
trained_file = data.lbp_frontal_face_cascade_filename()
#initialize the detector cascade
detector = Cascade(trained_file)
#detect the faces
detected = detector.detect_multi_scale(img=image, scale_factor=1.2, step_ratio=1, min_size=(20,20), max_size=(100,100))
#to display detected faces
showDetectedFaces(image,detected)
'''
To extract the face rectangle from the image using
the coordinates of the detected
'''
def getFace(d):
#X & Y are the starting points of the face rectangle
x,y = d['r'], d['c']
you want to have a really precise detector, you will have to train it yourself
using `OpenCV train cascade utility
<https://docs.opencv.org/doc/user_guide/ug_traincascade.html>`_.
"""
from skimage import data
from skimage.feature import Cascade
import matplotlib.pyplot as plt
from matplotlib import patches
# Load the trained file from the module root.
trained_file = data.lbp_frontal_face_cascade_filename()
# Initialize the detector cascade.
detector = Cascade(trained_file)
img = data.astronaut()
detected = detector.detect_multi_scale(img=img,
scale_factor=1.2,
step_ratio=1,
min_size=(60, 60),
max_size=(123, 123))
plt.imshow(img)
img_desc = plt.gca()
plt.set_cmap('gray')
for patch in detected:
# show the two images
sw_utils.functions.show_two_images(
test_img_cv,
test_img_gray,
cmap2='gray',
title='Image 1 opencv, Image 2 matplotlib')
my_detected = None
if control[1]:
"""
This is a follow up section to demonstrates taking the image from previous code block, and cropping the face
image
"""
# create a detector from Cascade
my_detector = Cascade(
os.path.join(WORK_ROOT, 'RES', 'FD', 'lbpcascade_frontalface.xml'))
# detecting face in an image
my_detected = my_detector.detect_multi_scale(img=test_img_cv,
scale_factor=1.2,
step_ratio=1,
min_size=(60, 60),
max_size=(123, 123))
img_desc = plt.gca()
plt.set_cmap('gray')
roi_color = []
for face in my_detected:
roi_color.append(test_img_cv[face['c']:face['c'] + face['height'],
face['r']:face['r'] + face['width']])
plt.imshow(image)
img_desc = plt.gca()
plt.set_cmap('rainbow')
plt.axis('off')
for patch in detected:
img_desc.add_patch(patches.Rectangle((patch['c'], patch['r']), patch['width'], patch['height'], fill=False, color='r', linewidth=2))
plt.show()
'''
#load the trained file from the module root
trained_file = data.lbp_frontal_face_cascade_filename()
#initialize the detector cascade
detector = Cascade(trained_file)
while (video_capture.isOpened()):
ret, image = video_capture.read()
image = cv2.resize(image,(640,480))
#if not ret:
# break
#detect the faces
detected = detector.detect_multi_scale(img=image, scale_factor=1.3, step_ratio=1.2, min_size=(30,30), max_size=(600,450))
#to display detected faces
showDetectedFaces(image,detected)
#Break the loop when user hits 'esc' key