def __init__(self,
p_model_path="fid/mtcnn/mtcnn_checkpoints/pnet_epoch.pt",
r_model_path="fid/mtcnn/mtcnn_checkpoints/rnet_epoch.pt",
o_model_path="fid/mtcnn/mtcnn_checkpoints/onet_epoch.pt",
use_cuda=1,
min_face_size=15):
pnet, rnet, onet = create_mtcnn_net(p_model_path, r_model_path,
o_model_path, use_cuda)
self.detector = MtcnnDetector(pnet=pnet,
rnet=rnet,
onet=onet,
min_face_size=min_face_size)
def main(csv_path, img_path):
pnet, rnet, onet = create_mtcnn_net(
p_model_path="fid/mtcnn/mtcnn_checkpoints/pnet_epoch.pt",
r_model_path="fid/mtcnn/mtcnn_checkpoints/rnet_epoch.pt",
o_model_path="fid/mtcnn/mtcnn_checkpoints/onet_epoch.pt",
use_cuda=True)
mtcnn_detector = MtcnnDetector(pnet=pnet,
rnet=rnet,
onet=onet,
min_face_size=24)
mobileFace = mobile_face_model(
"fid/InsightFace_Pytorch/facenet_checkpoints/model_ir_se50.pth")
labels, old_features = get_data(csv_path)
img = cv2.imread(img_path)
print(get_names(mtcnn_detector, mobileFace, labels, old_features, img))
def one_manImg(
img_dir,
csv_path,
model_path="fid/InsightFace_Pytorch/facenet_checkpoints/model_ir_se50.pth"
):
pnet, rnet, onet = create_mtcnn_net(
p_model_path="fid/mtcnn/mtcnn_checkpoints/pnet_epoch.pt",
r_model_path="fid/mtcnn/mtcnn_checkpoints/rnet_epoch.pt",
o_model_path="fid/mtcnn/mtcnn_checkpoints/onet_epoch.pt",
use_cuda=True)
mtcnn_detector = MtcnnDetector(pnet=pnet,
rnet=rnet,
onet=onet,
min_face_size=24)
mobileFace = mobile_face_model(model_path)
img_paths = os.listdir(img_dir)
if not os.path.exists(csv_path):
print('创建数据库')
with open(csv_path, 'w', encoding='UTF-8') as file_csv:
writer = csv.writer(file_csv)
header = ['Features%d' % x for x in range(512)]
header.insert(0, 'Name')
writer.writerow(header)
with open(csv_path, 'r', encoding='UTF-8') as file_csv:
reader = csv.reader(file_csv)
names = [row[0] for row in reader][1:]
with open(csv_path, 'a+', encoding='UTF-8') as file_csv:
writer = csv.writer(file_csv)
for img_path in img_paths:
label = os.path.splitext(img_path)[0]
if label in names:
print('%s的数据已存在,跳过' % label)
else:
img_path = os.path.join(img_dir, img_path)
image = cv2.imread(img_path)
faces, _ = detect_face(mtcnn_detector, image)
if len(faces) == 1:
print('正在输入:', label)
# cv2.imshow('demo', faces[0])
# cv2.waitKey(0)
features = get_faceFeatures(mobileFace, faces[0])[0]
content = np.append(label, features)
writer.writerow(content)
else:
print('%s图片不符合要求' % img_path)
def make_model():
yolo = yolo_model()
reid = FeatureExtractor(
model_name='osnet_x1_0',
model_path=
'pid/deep_person_reid/checkpoints/osnet_x1_0_market_256x128_amsgrad_ep150_stp60_lr0.0015_b64_fb10_softmax_labelsmooth_flip.pth',
verbose=False)
pnet, rnet, onet = create_mtcnn_net(
p_model_path="fid/mtcnn/mtcnn_checkpoints/pnet_epoch.pt",
r_model_path="fid/mtcnn/mtcnn_checkpoints/rnet_epoch.pt",
o_model_path="fid/mtcnn/mtcnn_checkpoints/onet_epoch.pt",
use_cuda=True)
mtcnn_detector = MtcnnDetector(pnet=pnet,
rnet=rnet,
onet=onet,
min_face_size=24)
mobileFace = mobile_face_model(
"fid/InsightFace_Pytorch/facenet_checkpoints/model_ir_se50.pth")
return yolo, reid, mtcnn_detector, mobileFace
if use_cuda:
images = images.cuda()
normal_features = []
with torch.no_grad():
features = faceModel(images)
for i in range(0, len(features), 2):
normal_feature = (features[i] + features[i + 1])
normal_features.append(l2_norm(normal_feature).cpu().detach().numpy())
return normal_features
if __name__ == '__main__':
pnet, rnet, onet = create_mtcnn_net(
p_model_path="fid/mtcnn/mtcnn_checkpoints/pnet_epoch.pt",
r_model_path="fid/mtcnn/mtcnn_checkpoints/rnet_epoch.pt",
o_model_path="fid/mtcnn/mtcnn_checkpoints/onet_epoch.pt",
use_cuda=True)
mtcnn_detector = MtcnnDetector(pnet=pnet,
rnet=rnet,
onet=onet,
min_face_size=24)
faceModel = mobile_face_model()
img = cv2.imread("/home/supermc/Pictures/IZTY.png")
faces, _ = detect_face(mtcnn_detector, img)
for id, face in enumerate(faces):
image_path = os.path.join("data/face_with_name", chr(id + 65) + ".png")
# cv2.imwrite(image_path, face)
# cv2.imshow('demo', face)
# cv2.waitKey(0)
features = get_faceFeatures(faceModel, faces)