def modify_frame(frame, clf, clf_th):
""" Plot a rectangle and a label if a face is identified in the frame
Arguments:
----------
frame:
type: numpy.array
info: array of frame pixels
clf:
type: FaceRecognizer object
info: trained classifier to identify faces
clf_th:
type: int / float
info: threshold to identify a face as 'Unknown'
Returns:
----------
frame:
type: numpy.array
info: array of frame pixels (may be modified)
"""
for face, coords in check_faces(frame):
face = normalize_face(face)
label = clf.predict(face, clf_th)
frame = draw_rect(frame, coords)
frame = draw_text(frame, label, coords)
return frame
def create_dataset(query, pics_num, search_engine=SEARCH_ENGINE):
""" Downloads, transforms and stores pictures given a query and an engine
Arguments:
----------
query:
type: string
info: query that fill be plot into the search engine
num:
type: int
info: number of pictures to obtain
search_engine:
type: dict (optional)
info: properties of the search engine to use. Keys:
- domain (string)
- path (string)
- params (dict)
"""
# Unwrapping search engine properties
domain = search_engine['domain']
path = search_engine['path']
params = search_engine['params']
params['q'] = query
stored = 0
# Until the number of pictures is not reached
while stored < pics_num:
if path:
url = build_url(domain, path, params)
page = get_page(url)
path = get_next_page(page, 'fl')
params = None
# Each image is saved if a face is detected
for image in get_images(page, url):
for face, _ in check_faces(image):
# Normalizes and stores the image
face = Image.fromarray(normalize_face(face))
save_image(image=face,
output_folder=query.replace(' ', '_'),
output_name=str(stored))
stored += 1
if stored == pics_num:
break