def run_job(params):
'''
Pre process image.
'''
directory = params[0]
root_path = params[1]
before_crop_resize = params[2]
image_size = params[3]
# Read and crop image.
image_files = Files(directory + "/")
directory = directory.split("/")
directory = directory[len(directory) - 1]
for image_file in image_files.paths:
image = Image(image_file)
image = crop_face(image, image_size)
if image.width > before_crop_resize:
image = image.resize(before_crop_resize)
# Save image.
file_name = image_file.split("/")
file_name = file_name[len(file_name) - 1]
image.save(root_path + directory + "/" + file_name)
def monitor(self, interval=0.5, time_span=10, threshold=1e-10, gray_scale=True):
""" During the past `time_span` seconds if there're less than
`threshold` * `time_span` pictures that's me, do something
"""
lookback = np.ones([int(time_span / interval)])
for image in self.cctv(interval, gray_scale):
faces = get_faces(image)
if faces is None:
print('No face detected in camera.')
continue
# get the face area
x, y, w, h = faces[0]
face = crop_face(image, x, w, y, h)
# resizing image
face = cv2.resize(face, (64, 64))
lookback = np.roll(lookback, shift=-1)
identity = whoisthis(face)
print(identity)
lookback[-1] = identity
if np.mean(lookback) < threshold:
show_image()
send_message(r'[Alert] Someone is look at your screen!')
raise ValueError('This is not me!')
def to_hdf5(wrapper, save_as=None):
if save_as is None:
save_as = wrapper.dataset_name.replace(' ', '_') + '.h5'
f = None
print wrapper.dataset_name
try:
f = tables.openFile(save_as, mode='w')
train_group = f.createGroup('/', 'train', 'train set')
test_group = f.createGroup('/', 'test', 'test set')
img_groups = (train_group, None)
test_idx = None
dsets = [range(len(wrapper)), None]
if wrapper.get_standard_train_test_splits() is not None:
dsets[0], dsets[1] = wrapper.get_standard_train_test_splits()
img_groups = (img_groups[0], test_group)
for some_idx, testing in enumerate(dsets):
if testing is None:
break
img_group = img_groups[some_idx]
dset = dsets[some_idx]
img_table = f.createCArray(img_group, 'img', tables.StringAtom(itemsize=1), shape=(len(dset), 96*96*3))
label_table = f.createCArray(img_group, 'label', tables.Float64Atom(), shape=(len(dset), len(keypoints_names), 2))
for i, _ in enumerate(dset):
img, label = crop_face(PIL.Image.open(wrapper.get_original_image_path(i)),
wrapper.get_bbox(i),
wrapper.get_eyes_location(i),
wrapper.get_keypoints_location(i))
img_table[i, :] = [x for x in img.tostring()]
#NOTE: will be stored as cstring, so it is mandatory to store as
#strings of size 1 (as \0 would break the string)
for name in keypoints_names:
if name in label:
point = label[name]
label_table[i, keypoints_names.index(name), :] = [point[0], point[1]]
else:
label_table[i, keypoints_names.index(name), :] = [-1, -1]
finally:
if f is not None:
f.close()
def subtract_neutral_face(img_path, dst='Same', flag='diff'):
# img_path: Path to jpg image, of expression face.
# dst: Destination to save cropped image.
# If 'same', will perform in-place operation. Or else, saved in relative location to img_path
# flag: Determine type of thresholding.
# If 'diff', return difference between neutral and expression face.
# If 'thre', return binary threshold of the difference.
# If 'ath', return adaptive threshold of the difference.
folder_path, filename = os.path.split(img_path)
if dst == 'Same':
dst_path = img_path
else:
dst_path = os.path.join(folder_path, dst)
# Extract and resize neutral face
nf_path = get_neutral_face_path(img_path)
nf_img = cv2.imread(nf_path, 0)
nf_img_crop = crop_face(nf_path, ret=True) #return 3D np array
nf_img_crop_rs = cv2.resize(nf_img_crop, (224, 224))
# Read and resize expression face
img = cv2.imread(img_path, 0)
img_rs = cv2.resize(img, (224, 224))
# Find absolute difference
diff = cv2.subtract(img_rs, nf_img_crop_rs)
retval, threshold = cv2.threshold(diff, 10, 255, cv2.THRESH_BINARY)
ath = cv2.adaptiveThreshold(diff, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 3)
# Find relative difference
#diff2 = img_rs - nf_img_crop_rs
#retval2, threshold2 = cv2.threshold(diff2, 50, 255, cv2.THRESH_BINARY)
#ath2 = cv2.adaptiveThreshold(diff2,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY,11,3)
if flag == 'diff':
return cv2.imwrite(dst_path, diff)
elif flag == 'thre':
return cv2.imwrite(dst_path, threshold)
elif flag == 'ath':
return cv2.imwrite(dst_path, ath)
def to_hdf5(wrapper, save_as=None):
if save_as is None:
save_as = wrapper.dataset_name.replace(' ', '_') + '.h5'
keypoints_to_id = {
'left_eyebrow_inner_end': 1,
'left_eyebrow_outer_end': 27,
'mouth_top_lip_bottom': 2,
'right_ear_canal': 3,
'face_right': 52,
'right_ear_top': 4,
'right_ear_center': 29,
'mouth_top_lip': 5,
'right_eye_center_top': 25,
'mouth_bottom_lip_top': 6,
'right_eyebrow_center': 7,
'face_center': 31,
'chin_left': 8,
'right_eyebrow_center_top': 33,
'face_left': 53,
'left_eyebrow_center': 34,
'right_eye_pupil': 35,
'right_nostril': 42,
'mouth_left_corner': 37,
'left_eye_center_bottom': 38,
'nose_tip': 9,
'nostrils_center': 18,
'left_eyebrow_center_top': 10,
'left_eye_outer_corner': 11,
'mouth_right_corner': 41,
'right_eye_inner_corner': 32,
'left_nostril': 48,
'right_eye_center': 43,
'right_ear': 12,
'left_ear_top': 28,
'mouth_bottom_lip': 13,
'left_eye_center': 14,
'left_mouth_outer_corner': 15,
'right_eyebrow_outer_end': 45,
'left_eye_center_top': 16,
'left_ear_center': 17,
'left_eye_pupil': 46,
'left_eyebrow_center_bottom': 39,
'right_ear_bottom': 36,
'right_eyebrow_inner_end': 26,
'right_eye_center_bottom': 20,
'chin_center': 21,
'left_eye_inner_corner': 22,
'mouth_center': 47,
'right_mouth_outer_corner': 23,
'left_ear_bottom': 24,
'nose_center_top': 30,
'right_eyebrow_center_bottom': 49,
'left_ear_canal': 50,
'right_eye_outer_corner': 19,
'left_ear': 51,
'chin_right': 44
}
f = None
print wrapper.dataset_name
try:
f = tables.openFile(save_as, mode='w')
train_group = f.createGroup('/', 'train', 'train set')
test_group = f.createGroup('/', 'test', 'test set')
train_img_group = f.createGroup('/train', 'data', 'data group')
train_label_group = f.createGroup('/train', 'label', 'label group')
test_img_group = f.createGroup('/test', 'data', 'data group')
test_label_group = f.createGroup('/test', 'label', 'label group')
img_groups = (train_img_group, None)
label_groups = (train_label_group, None)
test_idx = None
dsets = [range(len(wrapper)), None]
if wrapper.get_standard_train_test_splits() is not None:
dsets[0], dsets[1] = wrapper.get_standard_train_test_splits()
img_groups = (img_groups[0], test_img_group)
label_groups = (label_groups[0], test_label_group)
for some_idx, testing in enumerate(dsets):
if testing is None:
break
img_group = img_groups[some_idx]
label_group = label_groups[some_idx]
dset = dsets[some_idx]
img_table = f.createTable(img_group, 'img', ImgStruct,
'image data')
label_table = f.createTable(label_group, 'label', LabelStruct,
'target data')
img_row = img_table.row
label_row = label_table.row
for i in dset:
img, label = crop_face(
PIL.Image.open(wrapper.get_original_image_path(i)),
wrapper.get_bbox(i), wrapper.get_eyes_location(i),
wrapper.get_keypoints_location(i))
img_row['idx'] = i
img_row['data'] = img.tostring()
img_row.append()
for name, point in label.iteritems():
the_id = 0
if name in keypoints_to_id:
the_id = keypoints_to_id[name]
label_row['idx'] = i
label_row['name'] = the_id
label_row['col'] = point[0]
label_row['row'] = point[1]
label_row.append()
img_table.flush()
label_table.flush()
finally:
if f is not None:
f.close()
def to_hdf5(wrapper, save_as=None):
if save_as is None:
save_as = wrapper.dataset_name.replace(' ', '_') + '.h5'
keypoints_to_id = {'left_eyebrow_inner_end': 1, 'left_eyebrow_outer_end': 27, 'mouth_top_lip_bottom': 2, 'right_ear_canal': 3, 'face_right': 52, 'right_ear_top': 4, 'right_ear_center': 29, 'mouth_top_lip': 5, 'right_eye_center_top': 25, 'mouth_bottom_lip_top': 6, 'right_eyebrow_center': 7, 'face_center': 31, 'chin_left': 8, 'right_eyebrow_center_top': 33, 'face_left': 53, 'left_eyebrow_center': 34, 'right_eye_pupil': 35, 'right_nostril': 42, 'mouth_left_corner': 37, 'left_eye_center_bottom': 38, 'nose_tip': 9, 'nostrils_center': 18, 'left_eyebrow_center_top': 10, 'left_eye_outer_corner': 11, 'mouth_right_corner': 41, 'right_eye_inner_corner': 32, 'left_nostril': 48, 'right_eye_center': 43, 'right_ear': 12, 'left_ear_top': 28, 'mouth_bottom_lip': 13, 'left_eye_center': 14, 'left_mouth_outer_corner': 15, 'right_eyebrow_outer_end': 45, 'left_eye_center_top': 16, 'left_ear_center': 17, 'left_eye_pupil': 46, 'left_eyebrow_center_bottom': 39, 'right_ear_bottom': 36, 'right_eyebrow_inner_end': 26, 'right_eye_center_bottom': 20, 'chin_center': 21, 'left_eye_inner_corner': 22, 'mouth_center': 47, 'right_mouth_outer_corner': 23, 'left_ear_bottom': 24, 'nose_center_top': 30, 'right_eyebrow_center_bottom': 49, 'left_ear_canal': 50, 'right_eye_outer_corner': 19, 'left_ear': 51, 'chin_right': 44}
f = None
print wrapper.dataset_name
try:
f = tables.openFile(save_as, mode='w')
train_group = f.createGroup('/', 'train', 'train set')
test_group = f.createGroup('/', 'test', 'test set')
train_img_group = f.createGroup('/train', 'data', 'data group')
train_label_group = f.createGroup('/train', 'label', 'label group')
test_img_group = f.createGroup('/test', 'data', 'data group')
test_label_group = f.createGroup('/test', 'label', 'label group')
img_groups = (train_img_group, None)
label_groups = (train_label_group, None)
test_idx = None
dsets = [range(len(wrapper)), None]
if wrapper.get_standard_train_test_splits() is not None:
dsets[0], dsets[1] = wrapper.get_standard_train_test_splits()
img_groups = (img_groups[0], test_img_group)
label_groups = (label_groups[0], test_label_group)
for some_idx, testing in enumerate(dsets):
if testing is None:
break
img_group = img_groups[some_idx]
label_group = label_groups[some_idx]
dset = dsets[some_idx]
img_table = f.createTable(img_group, 'img', ImgStruct, 'image data')
label_table = f.createTable(label_group, 'label', LabelStruct, 'target data')
img_row = img_table.row
label_row = label_table.row
for i in dset:
img, label = crop_face(PIL.Image.open(wrapper.get_original_image_path(i)),
wrapper.get_bbox(i),
wrapper.get_eyes_location(i),
wrapper.get_keypoints_location(i))
img_row['idx'] = i
img_row['data'] = img.tostring()
img_row.append()
for name, point in label.iteritems():
the_id = 0
if name in keypoints_to_id:
the_id = keypoints_to_id[name]
label_row['idx'] = i
label_row['name'] = the_id
label_row['col'] = point[0]
label_row['row'] = point[1]
label_row.append()
img_table.flush()
label_table.flush()
finally:
if f is not None:
f.close()
import crop_face as cf
import cfal as alf
# parameters for loading data and images
prototxt = 'ckpt_/deploy.prototxt.txt'
weights = 'ckpt_/res10_300x300_ssd_iter_140000.caffemodel'
# loading models
face_detection = load_detection_model(prototxt, weights)
filename = "filepath_of_video"
dirName = "dir_name for cropped_images"
# for detection and saved the crop the detected face
a = cf.crop_face(filename, face_detection, dirName)
# for detection and align faces and saved the crop the detected face
# a = afl.crop_face(filename, face_detection, dirName)
print("Done", a)
def to_hdf5(wrapper, save_as=None):
if save_as is None:
save_as = wrapper.dataset_name.replace(' ', '_') + '.h5'
f = None
print wrapper.dataset_name
try:
f = tables.openFile(save_as, mode='w')
train_group = f.createGroup('/', 'train', 'train set')
test_group = f.createGroup('/', 'test', 'test set')
img_groups = (train_group, None)
test_idx = None
dsets = [range(len(wrapper)), None]
if wrapper.get_standard_train_test_splits() is not None:
dsets[0], dsets[1] = wrapper.get_standard_train_test_splits()
img_groups = (img_groups[0], test_group)
for some_idx, testing in enumerate(dsets):
if testing is None:
break
img_group = img_groups[some_idx]
dset = dsets[some_idx]
img_table = f.createCArray(img_group,
'img',
tables.StringAtom(itemsize=1),
shape=(len(dset), 96 * 96 * 3))
label_table = f.createCArray(img_group,
'label',
tables.Float64Atom(),
shape=(len(dset),
len(keypoints_names), 2))
for i, _ in enumerate(dset):
img, label = crop_face(
PIL.Image.open(wrapper.get_original_image_path(i)),
wrapper.get_bbox(i), wrapper.get_eyes_location(i),
wrapper.get_keypoints_location(i))
img_table[i, :] = [x for x in img.tostring()]
#NOTE: will be stored as cstring, so it is mandatory to store as
#strings of size 1 (as \0 would break the string)
for name in keypoints_names:
if name in label:
point = label[name]
label_table[i, keypoints_names.index(name), :] = [
point[0], point[1]
]
else:
label_table[i,
keypoints_names.index(name), :] = [-1, -1]
finally:
if f is not None:
f.close()
