def profile_face_clf(img_lst, cascade_file, scale_factor, min_neighbors, min_size):
"""Classifies face based on profile image
:param img_lst: list of image objects with profile image
:param cascade_file, scale_factor, min_neighbors, min_size: tuning parameters
:return pred_lst: list of predictions (face or not)
"""
pred_lst = []
i = 0
for img_obj in img_lst:
image = img_obj["face_img"]
face_detector = FaceDetector(cascade_file, scale_factor, min_neighbors, min_size)
face_detector_processor = FaceDetectorProcessor()
face_detector_processor.detector = face_detector
cropped_face, face_in_square = face_detector_processor.process_image(image)
i += 1
print "Inserting prediction result ", i
pred_lst.append(face_in_square)
return pred_lst
def profile_face_clf(img_lst, cascade_file, scale_factor, min_neighbors,
min_size):
"""Classifies face based on profile image
:param img_lst: list of image objects with profile image
:param cascade_file, scale_factor, min_neighbors, min_size: tuning parameters
:return pred_lst: list of predictions (face or not)
"""
pred_lst = []
i = 0
for img_obj in img_lst:
image = img_obj['face_img']
face_detector = FaceDetector(cascade_file, scale_factor, min_neighbors,
min_size)
face_detector_processor = FaceDetectorProcessor()
face_detector_processor.detector = face_detector
cropped_face, face_in_square = face_detector_processor.process_image(
image)
i += 1
print "Inserting prediction result ", i
pred_lst.append(face_in_square)
return pred_lst
def classify_profile_photo(self, image_metadata):
"""Classifies the face in a profile photo
Uses the CropImageProcessor to crop the image, passes it on to
the image_processor then use the clf to classify the image
:param image_metadata: metadata of the image
:return: classification result ([] if 'Content not found')
"""
url = image_metadata['user']['profile_picture']
imageFile = requests.get(url).content
if imageFile == 'Content not found':
return []
fout = open('temp.jpg', 'w')
fout.write(imageFile)
fout.close()
image = cv2.imread('temp.jpg')
height, width, channels = image.shape
photo_id = url[url.rindex('https://') + 9:].replace('/', '@')
clf_results = [] #list of classification results
user_id = image_metadata['user']['id']
#####Optimized parameters after model tuning
face_detector = FaceDetector('haarcascade_frontalface_default.xml',
1.1, 1, (20, 20))
face_detector_processor = FaceDetectorProcessor()
face_detector_processor.detector = face_detector
self._image_processor = face_detector_processor
#####
cropped_face, face_in_square = self._image_processor.process_image(
image)
try:
if face_in_square:
#stores it in directory
fname = self._image_processor.save_image(image=cropped_face,
user_id=user_id,
photo_id=photo_id)
input_image = caffe.io.load_image(fname)
clf_results.append((user_id, photo_id, self._clf(input_image)))
except Exception, e:
print str(e)
def classify_profile_photo(self, image_metadata):
"""Classifies the face in a profile photo
Uses the CropImageProcessor to crop the image, passes it on to
the image_processor then use the clf to classify the image
:param image_metadata: metadata of the image
:return: classification result ([] if 'Content not found')
"""
url = image_metadata['user']['profile_picture']
imageFile = requests.get(url).content
if imageFile == 'Content not found':
return []
fout = open('temp.jpg', 'w')
fout.write(imageFile)
fout.close()
image = cv2.imread('temp.jpg')
height, width, channels = image.shape
photo_id = url[url.rindex('https://')+9:].replace('/', '@')
clf_results = [] #list of classification results
user_id = image_metadata['user']['id']
#####Optimized parameters after model tuning
face_detector = FaceDetector('haarcascade_frontalface_default.xml',
1.1, 1, (20,20))
face_detector_processor = FaceDetectorProcessor()
face_detector_processor.detector = face_detector
self._image_processor = face_detector_processor
#####
cropped_face, face_in_square = self._image_processor.process_image(image)
try:
if face_in_square:
#stores it in directory
fname = self._image_processor.save_image(image=cropped_face,
user_id=user_id,
photo_id=photo_id)
input_image = caffe.io.load_image(fname)
clf_results.append((user_id, photo_id, self._clf(input_image)))
except Exception, e:
print str(e)
def tagged_face_clf(img_lst, cascade_file, scale_factor, min_neighbors,
min_size, crop_radius):
"""Classifies face based on full image and crop position
:param img_lst: list of image objects with full image and crop
position information
:param cascade_file, scale_factor, min_neighbors, min_size,
crop_radius: tuning parameters
:return pred_lst: list of predictions (face or not)
"""
pred_lst = []
i = 0
for img_obj in img_lst:
image = img_obj['full_img']
if image == None:
face_in_square = None
else:
pos_x = img_obj['crop_pos'][0]
pos_y = img_obj['crop_pos'][1]
height, width, channels = image.shape
x_coord = float(pos_x) * width
y_coord = float(pos_y) * height
crop_processor = CropImageProcessor((x_coord, y_coord),
tagged_radius=crop_radius)
cropped_square = crop_processor.process_image(image)
face_detector = FaceDetector(cascade_file, scale_factor,
min_neighbors, min_size)
face_detector_processor = FaceDetectorProcessor()
face_detector_processor.detector = face_detector
cropped_face, face_in_square = face_detector_processor.process_image(
cropped_square)
i += 1
print "Inserting prediction result", i
pred_lst.append(face_in_square)
return pred_lst
def tagged_face_clf(img_lst, cascade_file, scale_factor, min_neighbors, min_size, crop_radius):
"""Classifies face based on full image and crop position
:param img_lst: list of image objects with full image and crop
position information
:param cascade_file, scale_factor, min_neighbors, min_size,
crop_radius: tuning parameters
:return pred_lst: list of predictions (face or not)
"""
pred_lst = []
i = 0
for img_obj in img_lst:
image = img_obj["full_img"]
if image == None:
face_in_square = None
else:
pos_x = img_obj["crop_pos"][0]
pos_y = img_obj["crop_pos"][1]
height, width, channels = image.shape
x_coord = float(pos_x) * width
y_coord = float(pos_y) * height
crop_processor = CropImageProcessor((x_coord, y_coord), tagged_radius=crop_radius)
cropped_square = crop_processor.process_image(image)
face_detector = FaceDetector(cascade_file, scale_factor, min_neighbors, min_size)
face_detector_processor = FaceDetectorProcessor()
face_detector_processor.detector = face_detector
cropped_face, face_in_square = face_detector_processor.process_image(cropped_square)
i += 1
print "Inserting prediction result", i
pred_lst.append(face_in_square)
return pred_lst
def __init__(self, clf, image_processor=FaceDetectorProcessor()):
self._image_processor = image_processor
self._clf = clf
