def test_single(img):
if type(img) == str:
img = cv2.imread(img)
face_detector = FaceDetector(
name='mobilenet',
weight_path='../Retinaface/weights/mobilenet.pth',
device='cuda')
age_gender_detector = AgeGender(name='full',
weight_path='./weights/ShufflenetFull.pth',
device='cuda')
faces, boxes, scores, landmarks = face_detector.detect_align(img)
genders = None
ages = None
if len(faces.shape) > 1:
genders, ages = age_gender_detector.detect(faces)
for i, b in enumerate(boxes):
cv2.putText(img, f'{genders[i]},{ages[i]}',
(int(b[0]), int(b[1] - 10)), cv2.FONT_HERSHEY_SIMPLEX,
1.1, [0, 200, 0], 3)
cv2.rectangle(img, (int(b[0]), int(b[1])), (int(b[2]), int(b[3])),
(255, 0, 0), 3)
return img, genders, ages
default=True,
help="whether test time augmentation")
parser.add_argument('-m',
'--mobilenet',
default=False,
help="use mobilenet for backbone")
args = parser.parse_args()
conf = get_config(training=False)
conf.use_mobilfacenet = args.mobilenet
face_rec = face_learner(conf, inference=True)
face_rec.threshold = args.threshold
face_rec.model.eval()
face_detector = FaceDetector(
name='mobilenet',
weight_path='Retinaface/weights/mobilenet.pth',
device=conf.device)
age_gender_detector = AgeGender(
name='full',
weight_path='AgeGender/weights/ShufflenetFull.pth',
device=conf.device)
emotion_detector = EmotionDetector(
name='densnet121',
weight_path='FacialExpression/weights/densnet121.pth',
device=conf.device)
if args.update:
targets, names = update_facebank(conf,
face_rec.model,
face_detector,
tta=args.tta)
parser = argparse.ArgumentParser(description='take a picture')
parser.add_argument('--path',
'-p',
default='unknown',
type=str,
help='path of dir of person images')
args = parser.parse_args()
print('only a face in each image and all image from the same person')
dir_path = Path(args.path)
if not dir_path.is_dir():
exit('dir does not exists !!')
save_path = Path(f'data/facebank/{dir_path.name}')
if not save_path.exists():
save_path.mkdir()
# init detector
detector = FaceDetector(name='mobilenet',
weight_path='Retinaface/weights/mobilenet.pth',
device='cuda')
counter = 0
for img_path in dir_path.iterdir():
img = cv2.imread(str(img_path))
face = detector.detect_align(img)[0].cpu().numpy()
if len(face.shape) > 1:
save_name = f'{save_path}/{dir_path.name}_{counter}.jpg'
cv2.imwrite(save_name, face[0])
counter += 1
else:
print(img_path, 'in this image did not detect any face')
from Retinaface.Retinaface import FaceDetector
from FacialExpression.FaceExpression import EmotionDetector
import cv2
face_detector = FaceDetector(name='mobilenet',
weight_path='../Retinaface/weights/mobilenet.pth',
device='cuda',
face_size=(224, 224))
emotion_detector = EmotionDetector(name='densnet121',
weight_path='weights/densnet121.pth',
device='cuda')
vid = cv2.VideoCapture(0)
vid.set(3, 1280)
vid.set(4, 720)
while True:
ret, frame = vid.read()
faces, boxes, scores, landmarks = face_detector.detect_align(frame)
if len(faces.shape) > 1:
emotions, emo_probs = emotion_detector.detect_emotion(faces)
for b in boxes:
cv2.putText(frame, emotions[0], (int(b[0]), int(b[1]) - 5),
cv2.FONT_HERSHEY_SIMPLEX, 1.1, [0, 200, 0], 3)
cv2.rectangle(frame, (int(b[0]), int(b[1])),
(int(b[2]), int(b[3])), (255, 0, 0), 3)
for p in landmarks:
for i in range(5):
cv2.circle(frame, (p[i][0], p[i][1]), 3, (0, 255, 0), -1)
default='unknown',
type=str,
help='input the name of the recording person')
args = parser.parse_args()
save_path = Path('data/facebank') / args.name
if not save_path.exists():
save_path.mkdir()
# init camera
cap = cv2.VideoCapture(1)
cap.set(3, 1280)
cap.set(4, 720)
# init detector
detector = FaceDetector(name='resnet',
weight_path='Retinaface/weights/resnet50.pth',
device='cuda')
count = 4
while cap.isOpened():
_, frame = cap.read()
frame = cv2.putText(frame,
f'Press t to take {count} pictures, then finish...',
(10, 50), cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 100, 0), 3,
cv2.LINE_AA)
if cv2.waitKey(1) & 0xFF == ord('t'):
count -= 1
faces = detector.detect_align(frame)[0].cpu().numpy()
if len(faces.shape) > 1:
cv2.imwrite(f'{save_path}/{args.name}_{count}.jpg', faces[0])
if count <= 0: