def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 4))
bn = BatchNormalization()(input_tensor)
conv = Conv2D(3, (3, 3), padding='same', activation='relu')(bn)
base_model = ResNet34(include_top=False,
weights=None,
input_shape=(SIZE, SIZE, 3),
input_tensor=conv)
x = base_model.output
x = Conv2D(32, kernel_size=(1, 1), activation='relu')(x)
x = Flatten()(x)
x = Dropout(0.5)(x)
x = Dense(1024, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
res_img = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3))
offset = 2
for i, l in enumerate(base_model.layers[offset + 1:]):
l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
#bn = BatchNormalization()(input_tensor)
#conv = Conv2D(3,(3,3),padding='same',activation='relu')(bn)
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),
input_tensor=input_tensor)
enc_out, short_cuts = encoder(base_model)
x0 = GlobalAveragePooling2D()(squeeze_excite_block(enc_out))
x1 = GlobalAveragePooling2D()(squeeze_excite_block(short_cuts[0]))
x2 = GlobalAveragePooling2D()(squeeze_excite_block(short_cuts[1]))
x = Concatenate()([x0, x1, x2])
x = Dropout(0.3)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.3)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.3)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
#bn = BatchNormalization()(input_tensor)
#conv = Conv2D(3,(3,3),padding='same',activation='relu')(bn)
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),
input_tensor=input_tensor)
x = base_model.get_layer('stage3_unit1_relu1').output
#x = Dropout(0.5)(x)
x_map32 = Conv2D(28,
kernel_size=(1, 1),
activation='sigmoid',
name='map32')(x)
x_pooled = NoisyAndPooling2D()(x_map32)
output = Dense(n_out, activation='sigmoid', name='pred')(x_pooled)
model = Model(input_tensor, [output, x_map32])
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
#bn = BatchNormalization()(input_tensor)
#conv = Conv2D(3,(3,3),padding='same',activation='relu')(bn)
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),input_tensor=input_tensor)
x = base_model.output
x1 = Conv2D(32, kernel_size=(1, 1), activation='relu')(x)
x1 = Flatten()(x1)
x1 = Dropout(0.5)(x1)
x1 = Dense(1024, activation='relu')(x1)
x1 = Dropout(0.5)(x1)
predictions = Dense(n_out, activation='sigmoid',name = 'predictions')(x1)
x2 = decoder_block(x,5,512)
img_out = Dense(3, activation='sigmoid', name='img_out')(x2)
model = Model(input_tensor, [predictions,img_out])
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),input_tensor=input_tensor)
x = base_model.output
x1 = GlobalMaxPooling2D()(x)
x2 = GlobalAveragePooling2D()(x)
x = Concatenate()([x1,x2])
#x = BatchNormalization(axis=-1)(x)
x = Dropout(0.5)(x)
x = Dense(512, activation='relu')(x)
#x = BatchNormalization(axis=-1)(x)
x = Dropout(0.5)(x)
output = Dense(n_out)(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),
input_tensor=input_tensor)
enc_out, short_cuts = encoder(base_model)
x = GlobalAveragePooling2D()(enc_out)
#x = Conv2D(32, kernel_size=(1, 1), activation='relu')(x)
#x = Flatten()(x)
x = Dropout(0.5)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
red = Lambda(lambda x: x[:, :, :, :1])(input_tensor)
green = Lambda(lambda x: x[:, :, :, 1:2])(input_tensor)
blue = Lambda(lambda x: x[:, :, :, 2:])(input_tensor)
#bn = BatchNormalization()(input_tensor)
#conv = Conv2D(3,(3,3),padding='same',activation='relu')(bn)
red_base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3))
red_base_model.name = 'resnet34_red'
green_base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3))
green_base_model.name = 'resnet34_green'
blue_base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3))
blue_base_model.name = 'resnet34_blue'
x = Add()([
red_base_model(Concatenate(axis=-1)([red, red, red])),
green_base_model(Concatenate(axis=-1)([green, green, green])),
blue_base_model(Concatenate(axis=-1)([blue, blue, blue]))
])
x = Conv2D(256, (3, 3), activation='relu')(x)
x = GlobalMaxPooling2D()(x)
x = Dropout(0.5)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=input_shape)
#bn = BatchNormalization()(input_tensor)
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=input_shape,input_tensor=input_tensor)
x = base_model.output
x = Conv2D(32, kernel_size=(1, 1), activation='relu')(x)
x = Flatten()(x)
x = Dropout(0.5)(x)
x = Dense(1024, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
return model
def create_model(n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),
input_tensor=input_tensor)
enc_out, short_cuts = encoder(base_model)
x0 = GlobalAveragePooling2D()(enc_out)
x1 = GlobalAveragePooling2D()(short_cuts[0])
x2 = GlobalAveragePooling2D()(short_cuts[1])
x3 = GlobalAveragePooling2D()(short_cuts[2])
x4 = GlobalAveragePooling2D()(short_cuts[3])
x = Concatenate()([x0, x1, x2, x3, x4])
x = Dropout(0.5)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
up_left = Cropping2D(((0, 128), (0, 128)))(input_tensor)
up_left = UpSampling2D()(up_left)
up_right = Cropping2D(((128, 0), (0, 128)))(input_tensor)
up_right = UpSampling2D()(up_right)
mid = Cropping2D(((64, 64), (64, 64)))(input_tensor)
mid = UpSampling2D()(mid)
bot_left = Cropping2D(((0, 128), (128, 0)))(input_tensor)
bot_left = UpSampling2D()(bot_left)
bot_right = Cropping2D(((128, 0), (128, 0)))(input_tensor)
bot_right = UpSampling2D()(bot_right)
#bn = BatchNormalization()(input_tensor)
#conv = Conv2D(3,(3,3),padding='same',activation='relu')(bn)
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3))
up_left_out = top(base_model(up_left))
up_right_out = top(base_model(up_right))
mid_out = top(base_model(mid))
bot_left_out = top(base_model(bot_left))
bot_right_out = top(base_model(bot_right))
x = Concatenate()(
[up_left_out, up_right_out, mid_out, bot_left_out, bot_right_out])
x = Dropout(0.5)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3))
x = base_model(input_tensor)
x = Conv2D(256, (3, 3), activation='relu')(x)
x = GlobalMaxPooling2D()(x)
x = Dropout(0.5)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model
def create_model(input_shape, n_out):
input_tensor = Input(shape=(SIZE, SIZE, 3))
#bn = BatchNormalization()(input_tensor)
#conv = Conv2D(3,(3,3),padding='same',activation='relu')(bn)
base_model = ResNet34(include_top=False,
weights='imagenet',
input_shape=(SIZE, SIZE, 3),input_tensor=input_tensor)
x = base_model.output
x = GlobalAveragePooling2D()(x)
#x = Flatten()(x)
x = Dropout(0.5)(x)
x = Dense(256, activation='relu')(x)
x = Dropout(0.5)(x)
output = Dense(n_out, activation='sigmoid')(x)
model = Model(input_tensor, output)
# transfer imagenet weights
#res_img = ResNet34(include_top=False, weights='imagenet', input_shape=(SIZE, SIZE, 3))
#offset = 2
#for i, l in enumerate(base_model.layers[offset+1:]):
# l.set_weights(res_img.layers[i + 1].get_weights())
return model