def generate_dataset(root_face_images_folder_path,
mask_images_folder="mask_images"):
original_images_paths = list(
pathlib.Path(root_face_images_folder_path).glob('*'))
mask_images_paths = list(pathlib.Path(mask_images_folder).glob('*'))
mask_images = [Image.open(path) for path in mask_images_paths]
targets = {}
for i, img_path in enumerate(tqdm(original_images_paths)):
image_name = str(img_path).split(os.sep)[1]
image = Image.open(img_path).convert("RGB").resize((800, 800))
detected_faces_boxes, detected_faces_landmarks = detect_faces(
np.array(image), False)
if len(detected_faces_boxes) == 0:
continue
image.save("resized_images/" + image_name)
image = apply_face_mask(image, mask_images, detected_faces_landmarks)
image.save("generated_images/" + image_name)
targets[image_name] = {
"bbox": detected_faces_boxes,
}
# cv2.imshow("Output", cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR))
# cv2.waitKey(0)
with open("targets.json", "w", encoding="utf8") as outfile:
json.dump(targets, outfile)
def post(self):
args = parser.parse_args()
img_url = args['img']
response = req.get(img_url)
img = Image.open(BytesIO(response.content))
bounding_boxes = detect_faces(
img) # detect bboxes and landmarks for all faces in the image
width, height = img.size
#wnum:横向分割块数 hnum：纵向分割块数
wnum = int(args['ysplit'])
hnum = int(args['xsplit'])
w_step = width / wnum
h_step = height / hnum
flag = np.ones((hnum, wnum))
val = np.ones((hnum, wnum, 2))
res = []
for i in range(0, len(bounding_boxes)):
y = bounding_boxes[i][0] / (w_step + 0.001)
x = bounding_boxes[i][1] / (h_step + 0.001)
flag[int(x)][int(y)] += 1
val[int(x)][int(y)][0] += bounding_boxes[i][0]
val[int(x)][int(y)][1] += bounding_boxes[i][1]
for i in range(0, int(args['nums'])):
mpos = np.unravel_index(np.argmax(flag), flag.shape)
res.append([
val[mpos[0], mpos[1], 1] / np.max(flag) / height,
val[mpos[0], mpos[1], 0] / np.max(flag) / width
])
flag[mpos[0], mpos[1]] = 0
return res, 201
def align_face(img_path, resize_face_size=112):
img = Image.open(img_path)
bounding_boxes, landmarks = detect_faces(img)
scale = resize_face_size / 112.
reference = get_reference_facial_points(default_square = True) * scale
facial5points = [[landmarks[0][j], landmarks[0][j + 5]] for j in range(5)]
warped_face = warp_and_crop_face(np.array(img), facial5points, reference, crop_size=(resize_face_size, resize_face_size))
return warped_face
def image_gaze(img_dir):
for image in glob.glob(img_dir):
head, tail = os.path.split(image)
# stage1
img = Image.open(image) # modify the image path to yours
image = cv2.imread(image, cv2.IMREAD_COLOR)
bounding_boxes, landmarks = detect_faces(
img) # detect bboxes and landmarks for all faces in the image
h, w = image.shape[:2]
for i in range(len(bounding_boxes)):
#imgpts, modelpts, rotate_degree, nose = face_orientation(image, landmarks[i])
#random_color = tuple(np.random.random_integers(0, 255, size=3))
cv2.rectangle(
image, (int(bounding_boxes[i][0]), int(bounding_boxes[i][1])),
(int(bounding_boxes[i][2]), int(bounding_boxes[i][3])),
(0, 155, 255), 2)
headpose_center_cords = normalize_eyeCords(
int(bounding_boxes[i][0]), int(bounding_boxes[i][1]),
int(bounding_boxes[i][2]), int(bounding_boxes[i][3]), w, h)
px, py = detect_target(
image,
(headpose_center_cords[0][0], headpose_center_cords[0][1]))
draw_result(
image,
(headpose_center_cords[0][0], headpose_center_cords[0][1]),
(px, py), [0, 0, 255])
#stage2
result = detector.detect_faces(image)
if result != []:
for person in result:
#imgpts, modelpts, rotate_degree, nose = face_orientation(image,person['keypoints'])
#cv2.line(image, nose, tuple(imgpts[1].ravel()), (0, 255, 0), 2) # GREEN
#cv2.line(image, nose, tuple(imgpts[0].ravel()), (255, 0,), 2) # BLUE
#cv2.line(image, nose, tuple(imgpts[2].ravel()), (0, 0, 255), 2) # RED
confidence = person['confidence']
if (confidence > 0.90):
bounding_box = person['box']
keypoints = person['keypoints']
#commented cv2 drawing
#cv2.rectangle(image,(bounding_box[0], bounding_box[1]),(bounding_box[0] + bounding_box[2], bounding_box[1] + bounding_box[3]),(0, 155, 255),2)
#cv2.circle(image, (keypoints['left_eye']), 2, (0, 155, 255), 2)
#cv2.circle(image, (keypoints['right_eye']), 2, (0, 155, 255), 2)
leyecoords, reyecoords = enormalize(
keypoints['left_eye'], keypoints['right_eye'], w, h)
lpx, lpy = detect_target(
image, (leyecoords[0][0], leyecoords[0][1]))
rpx, rpy = detect_target(
image, (reyecoords[0][0], reyecoords[0][1]))
draw_result(image, (leyecoords[0][0], leyecoords[0][1]),
(lpx, lpy), [255, 0, 0])
draw_result(image, (reyecoords[0][0], reyecoords[0][1]),
(rpx, rpy), [0, 255, 255])
cv2.imwrite("./output/" + tail, image)
cv2.waitKey(0)
def get_central_face_attributes(full_path):
img = cv.imread(full_path, cv.IMREAD_COLOR)
bboxes, landmarks = detect_faces(img)
if len(landmarks) > 0:
i = select_significant_face(bboxes)
return [bboxes[i]], [landmarks[i]]
return None, None
def detect():
if request.method == 'POST':
print(request)
if 'file' not in request.files:
return jsonify({'status': "FAILURE", "message": "No file part found."})
file = request.files['file']
filename = secure_filename(file.filename)
file.save(os.path.join(app.config['DATA_FOLDER'], filename))
return detect_faces(filename)
def get_face_attributes(full_path):
try:
img = cv.imread(full_path, cv.IMREAD_COLOR)
bounding_boxes, landmarks = detect_faces(img)
if len(landmarks) > 0:
landmarks = [int(round(x)) for x in landmarks[0]]
return True, landmarks
except KeyboardInterrupt:
raise
except Exception as err:
print(err)
return False, None
def detect_face():
filename = 'face.jpg'
try:
print(request)
request_data = request.get_json()
file = request_data['file']
starter = file.find(',')
image_data = file[starter + 1:]
image_data = bytes(image_data, encoding="ascii")
im = Image.open(BytesIO(base64.b64decode(image_data)))
im.save(os.path.join(app.config['DATA_FOLDER'], filename))
except Exception as e:
print("Exception while detecting faces: ", e)
return jsonify({"status": "FAILURE", "message": "Unknown error while getting encoded file."})
return detect_faces(filename)
def take_images():
# initialize the camera
cam = cv2.VideoCapture(0) # 0 -> index of camera
log.info('Initialized camera capturing')
captured = True
while captured:
captured, img = cam.read()
time.sleep(sleep_time)
if captured:
# log.debug('Image captured')
if detector.detect_faces(img) or detector.detect_people(img):
log.info('Detected people')
cv2.imwrite(hierarchical_file(datetime.now(), detection=True), resize_image(img, people_scale_factor))
elif random.random() > 0.5:
cv2.imwrite(hierarchical_file(datetime.now(), detection=False), resize_image(img, no_people_scale_factor))
else:
log.error('No image captured')
sys.exit(1)
def get_all_face_attributes(full_path):
img = cv.imread(full_path, cv.IMREAD_COLOR)
bounding_boxes, landmarks = detect_faces(img)
return bounding_boxes, landmarks
def predict_image():
data = {'success': False}
if request.method == 'POST':
if request.files.get('image'):
f = request.files.get('image')
type = secure_filename(f.filename).split('.')[1]
if type not in ALLOWED_EXTENSIONS:
return 'Invalid type of file'
if f:
filename = os.path.join(app.config['UPLOAD_FOLDER'],
secure_filename(f.filename))
f.save(filename)
elif request.form['url']:
try:
url = request.form.get('url')
print(url)
f = urllib.request.urlopen(url)
filename = url.split('/')[-1]
filename = secure_filename(filename)
if filename:
filename = os.path.join(app.config['UPLOAD_FOLDER'],
filename)
f.save(filename)
except:
print('Cannot read image from url')
if filename:
fn = secure_filename(filename)[:-4]
min_side = 512
img = cv2.imread(filename)
size = img.shape
h, w = size[0], size[1]
if max(w, h) > min_side:
img_pad = process_image(img)
else:
img_pad = img
cv2.imwrite(
os.path.join(app.config['UPLOAD_FOLDER'], f'{fn}_resize.png'),
img_pad)
img = Image.open(
os.path.join(app.config['UPLOAD_FOLDER'], f'{fn}_resize.png'))
bounding_boxes, landmarks = detect_faces(
img) # detect bboxes and landmarks for all faces in the image
pic_face_detect = show_results(img, bounding_boxes,
landmarks) # visualize the results
pic_face_detect.save(
os.path.join(app.config['UPLOAD_FOLDER'],
f'{fn}_landmark.png'))
crop_size = 224
scale = crop_size / 112
reference = get_reference_facial_points(
default_square=True) * scale
for i in range(len(landmarks)):
facial5points = [[landmarks[i][j], landmarks[i][j + 5]]
for j in range(5)]
warped_face = warp_and_crop_face(np.array(img),
facial5points,
reference,
crop_size=(crop_size,
crop_size))
img_warped = Image.fromarray(warped_face)
pic_face_crop = img_warped.save(
os.path.join(UPLOAD_FOLDER, f'{fn}_{i}_crop.png'))
# face recognition
cleb_name = []
for i in range(len(landmarks)):
name = model.predict(
os.path.join(app.config['UPLOAD_FOLDER'],
f'{fn}_{i}_crop.png'))
cleb_name.append(name)
employeeList = []
for i in range(len(landmarks)):
for j in bounding_boxes:
face = {
"bounding_boxes": {
"top": j[0],
"right": j[1],
"left": j[2],
"bottom": j[3]
},
"landmark": landmarks[i],
"prediction": cleb_name[i],
"success": True
}
employeeList.append(face)
return jsonify(jsanitize(employeeList))
def detect_faces(*args, **kwargs):
"""wrapper for detector.detect_faces"""
with peek("./face.evoLVe.PyTorch/align"), torch.no_grad():
return detector.detect_faces(*args, **kwargs)
print("Welcome to the Facial Recognition doorbell!\n")
if not os.path.exists('database.db'):
print('We have detected this is your first time running this program! We will create a database for you!\n')
create_database.create_database()
running = True
while running:
user_input = input("You can start the doorbell by typing 1\n"
"You can record a new face by typing 2\n"
"You can exit by typing 3\n"
"---> ")
if user_input == "1":
detector.detect_faces()
cv2.destroyAllWindows()
elif user_input == "2":
record_faces.record_faces()
trainer.train_faces()
print("Face saved!")
elif user_input == "3":
running = False
else:
print("Invalid Input! Try again")
print("Thank you for using the Facial Recognition doorbell!")
def detect(photo):
detection = detector.detect_faces(photo['source'])
if 'tags' not in detection:
print detection, '\nExiting!'
return None #sys.exit()
return detection
from PIL import Image
from detector import detect_faces
import os, glob
path = "./data/AgeDB/*.jpg"
for img_file in glob.glob(path):
im = Image.open(img_file)
bounding_boxes, landmarks = detect_faces(im)
print("For file", img_file, "the bounding boxes are:", str(bounding_boxes))
crop_size = args.crop_size # specify size of aligned faces, align and crop with padding
scale = crop_size / 112.
reference = get_reference_facial_points(default_square = True) * scale
cwd = os.getcwd() # delete '.DS_Store' existed in the source_root
os.chdir(source_root)
os.system("find . -name '*.DS_Store' -type f -delete")
os.chdir(cwd)
if not os.path.isdir(dest_root):
os.mkdir(dest_root)
for subfolder in tqdm(os.listdir(source_root)):
if not os.path.isdir(os.path.join(dest_root, subfolder)):
os.mkdir(os.path.join(dest_root, subfolder))
for image_name in os.listdir(os.path.join(source_root, subfolder)):
print("Processing\t{}".format(os.path.join(source_root, subfolder, image_name)))
img = Image.open(os.path.join(source_root, subfolder, image_name))
try: # Handle exception
_, landmarks = detect_faces(img)
except Exception:
print("{} is discarded due to exception!".format(os.path.join(source_root, subfolder, image_name)))
continue
if len(landmarks) == 0: # If the landmarks cannot be detected, the img will be discarded
print("{} is discarded due to non-detected landmarks!".format(os.path.join(source_root, subfolder, image_name)))
continue
facial5points = [[landmarks[0][j], landmarks[0][j + 5]] for j in range(5)]
warped_face = warp_and_crop_face(np.array(img), facial5points, reference, crop_size=(crop_size, crop_size))
img_warped = Image.fromarray(warped_face)
img_warped.save(os.path.join(dest_root, subfolder, image_name.split('.')[0] + '.jpg'))
os.chdir(source_root)
os.system("find . -name '*.DS_Store' -type f -delete")
os.chdir(cwd)
if not os.path.isdir(dest_root):
os.mkdir(dest_root)
for subfolder in tqdm(os.listdir(source_root)):
if not os.path.isdir(os.path.join(dest_root, subfolder)):
os.mkdir(os.path.join(dest_root, subfolder))
for image_name in os.listdir(os.path.join(source_root, subfolder)):
if os.path.exists(os.path.join(dest_root, subfolder, image_name)):
continue
img = Image.open(os.path.join(source_root, subfolder, image_name))
try: # Handle exception
_, landmarks = detect_faces(img, gpu_id)
except Exception:
print("{} is discarded due to exception!".format(
os.path.join(source_root, subfolder, image_name)))
continue
if len(
landmarks
) == 0: # If the landmarks cannot be detected, the img will be discarded
print("{} is discarded due to non-detected landmarks!".format(
os.path.join(source_root, subfolder, image_name)))
continue
facial5points = [[landmarks[0][j], landmarks[0][j + 5]]
for j in range(5)]
warped_face = warp_and_crop_face(np.array(img),
facial5points,
reference,
os.system("find . -name '*.DS_Store' -type f -delete")
os.system("find . -name '*.ipynb_checkpoints' -prune -exec rm -rf {} \;")
os.chdir(cwd)
if not os.path.isdir(dest_root):
os.mkdir(dest_root)
for subfolder in tqdm(os.listdir(source_root)):
if not os.path.isdir(os.path.join(dest_root, subfolder)):
os.mkdir(os.path.join(dest_root, subfolder))
for image_name in os.listdir(os.path.join(source_root, subfolder)):
print("Processing\t{}".format(
os.path.join(source_root, subfolder, image_name)))
img = Image.open(os.path.join(source_root, subfolder, image_name))
try: # Handle exception
_, landmarks = detect_faces(img)
except Exception:
print("{} is discarded due to exception!".format(
os.path.join(source_root, subfolder, image_name)))
continue
if len(
landmarks
) == 0: # If the landmarks cannot be detected, the img will be discarded
print("{} is discarded due to non-detected landmarks!".format(
os.path.join(source_root, subfolder, image_name)))
continue
facial5points = [[landmarks[0][j], landmarks[0][j + 5]]
for j in range(5)]
warped_face = warp_and_crop_face(np.array(img),
facial5points,
reference,
uid = uid[0:len(uid)-5] + '@wicknicks'
if uid in cache:
print uid, 'has already been trained'
print 'Exiting'
sys.exit()
print 'Tagging', uid
if len(sys.argv) < 2: sys.exit()
links = open(sys.argv[1], 'r')
count = 0
for img in links.readlines():
img = img.strip();
detection = detector.detect_faces(img)
if 'tags' not in detection:
print detection
sys.exit()
if len(detection['tags']) != 1:
print "# of tags != ", img, len(detection['tags'])
print json.dumps(detection, sort_keys = True, indent = 2)
print
sys.exit()
tag = detection['tags'][0]['tid']
rsp = tagger.save_tag(uid, tag, img)
count += 1