def load_model(self, model_dir_path):
self.model = faceRecoModel(input_shape=(3, 96, 96))
print("Total Params:", self.model.count_params())
self.model.compile(optimizer='adam',
loss=triplet_loss,
metrics=['accuracy'])
load_weights_from_FaceNet(self.model, model_dir_path)
def load_weights(FRmodel) -> None:
"""Load weights from FaceNet model to avoid training cost.
Argeuments:
FRmodel -- """
FRmodel = faceRecoModel(input_shape=(3, 96, 96))
FRmodel.compile(optimizer='adam', loss=triplet_loss, metrics=['accuracy'])
load_weights_from_FaceNet(FRmodel)
def load_feature_model(model_file):
if os.path.exists(model_file):
# https://github.com/keras-team/keras/issues/5916
# import keras.losses
# keras.losses.custom_loss = triplet_loss
# FRmodel = load_model(model_file)
start = time.time()
model = load_model(model_file,
custom_objects={'triplet_loss': triplet_loss})
done = time.time()
print(f"model loaded in {done-start} seconds")
else:
model = faceRecoModel(input_shape=(3, model_width, model_height))
model.compile(optimizer='adam',
loss=triplet_loss,
metrics=['accuracy'])
load_weights_from_FaceNet(model)
model.save(model_file)
print("Total Params:", model.count_params())
return model
def main():
start_time = time.time()
np.set_printoptions(threshold=np.nan)
FRmodel = faceRecoModel(input_shape=(3, 96, 96))
print("Total Params:", FRmodel.count_params())
FRmodel.compile(optimizer='adam', loss=triplet_loss, metrics=['accuracy'])
load_weights_from_FaceNet(FRmodel)
# reate database
database = dict()
database["younes"] = img_to_encoding("images/younes.jpg", FRmodel)
database["kevin"] = img_to_encoding("images/kevin.jpg", FRmodel)
database["andrew"] = img_to_encoding("images/andrew.jpg", FRmodel)
database["sebastiano"] = img_to_encoding("images/sebastiano.jpg", FRmodel)
database["moi"] = img_to_encoding("images/moi3.jpg", FRmodel)
verify("images/camera_0.jpg", "younes", database, FRmodel)
print('')
print("---- camera_0 ----")
who_is_it("images/camera_0.jpg", database, FRmodel)
print('')
print("---- moi1 ----")
who_is_it("images/moi1.jpg", database, FRmodel, display=True)
print('')
print("---- moi2 ----")
who_is_it("images/moi2.jpg", database, FRmodel, display=True)
print('')
print("---- moi4 ----")
who_is_it("images/moi4.jpg", database, FRmodel, display=True)
print('')
print("---- moi5 ----")
who_is_it("images/moi5.jpg", database, FRmodel, display=True)
print('')
print("--- %s seconds ---" % (time.time() - start_time))
else:
print("it's " + str(identity) + ", the distance is " + str(min_dist))
return min_dist, identity
def resize_image(url, width, height):
with open(url, 'r+b') as f:
with Image.open(f) as image:
cover = resizeimage.resize_cover(image, [width, height])
cover.save(url, image.format)
if __name__ == '__main__':
K.set_image_data_format('channels_first')
FRmodel = faceRecoModel(input_shape=(3, 96, 96))
load_weights(FRmodel)
database = dict()
resize_image("/Users/markmc/Repos/RecogNet/faces/liis_01.jpg", 96, 96)
resize_image("/Users/markmc/Repos/RecogNet/faces/mark_01.jpg", 96, 96)
resize_image("/Users/markmc/Repos/RecogNet/faces/liis_02.jpg", 96, 96)
resize_image("/Users/markmc/Repos/RecogNet/faces/mark_02.jpg", 96, 96)
database["lisa"] = img_to_encoding(
"/Users/markmc/Repos/RecogNet/faces/liis_03.jpg", FRmodel)
database["mark"] = img_to_encoding(
"/Users/markmc/Repos/RecogNet/faces/mark_05.jpg", FRmodel)
def initModel():
model = faceRecoModel(input_shape=INPUT_SHAPE)
utils.load_weights_from_FaceNet(model)
return model
def myMainMethod():
FRmodel = inception_blocks_v2.faceRecoModel(input_shape=(3, 96, 96))
encoding = fr_utils.img_path_to_encoding("crop_frame.jpg", FRmodel)
print(encoding)
# import fr_utils as fr_utils
# import inception_blocks_v2 as inception_blocks_v2
import fr_utils
import inception_blocks_v2
# 3 つのライブラリを読み込む
# import importlib
# libs = ['fr_utils', 'inception_blocks_v2']
# for lib in libs:
# importlib.import_module(lib)
PADDING = 50
ready_to_detect_identity = True
#windows10_voice_interface = wincl.Dispatch("SAPI.SpVoice")
print( f"create faceRecModel...")
FRmodel = inception_blocks_v2.faceRecoModel(input_shape=(3, 96, 96))
print( f"create faceRecModel...done")
def triplet_loss(y_true, y_pred, alpha = 0.3):
"""
Implementation of the triplet loss as defined by formula (3)
Arguments:
y_pred -- python list containing three objects:
anchor -- the encodings for the anchor images, of shape (None, 128)
positive -- the encodings for the positive images, of shape (None, 128)
negative -- the encodings for the negative images, of shape (None, 128)
Returns:
loss -- real number, value of the loss
"""
def createmdl_img(self, input_shape):
self.FRmodel = inception_blocks_v2.faceRecoModel(input_shape)
print("model is created and Total Params:" +
str(self.FRmodel.count_params()))
return self.FRmodel
return min_dist, identity
def img_to_encoding(image_path, model):
img1 = cv2.imread(image_path, 1)
img = img1[..., ::-1]
img = np.around(np.transpose(img, (2, 0, 1)) / 255.0, decimals=12)
x_train = np.array([img])
encoding = model.predict_on_batch(x_train)
return encoding
# Main Code
FRmodel = ib.faceRecoModel(input_shape=(3, 93, 93))
us.load_weights_from_FaceNet(FRmodel)
database = {}
database["Negar"] = img_to_encoding("images/Negar1.jpg", FRmodel)
database["MySister"] = img_to_encoding("images/Nazanin.jpg", FRmodel)
print()
print('Verify if the new image is Negar:')
face_verification("images/Negar4.jpg", "Negar", database, FRmodel)
print('*********************************')
face_verification("images/Nazanin.jpg", "Negar", database, FRmodel)
print('*********************************')
print('Recognize who is in the picture:')