def Tiramisu(self, params):
"""
References:
Simon Jegou et al. The One Hundred Layers Tiramisu: Fully Convolutional Densenets for Semantic Segmentation.
"""
self.ids_inputs = params["INPUTS_IDS_MODEL"]
self.ids_outputs = params["OUTPUTS_IDS_MODEL"]
crop = params['IMG_CROP_SIZE']
image = Input(shape=tuple([crop[-1], None, None]),
name=self.ids_inputs[0])
init_filters = params['TIRAMISU_INIT_FILTERS']
growth_rate = params['TIRAMISU_GROWTH_RATE']
num_transition_blocks = params['TIRAMISU_N_TRANSITION_BLOCKS']
n_layers_down = params['TIRAMISU_N_LAYERS_DOWN']
n_layers_up = params['TIRAMISU_N_LAYERS_UP']
bottleneck_layers = params['TIRAMISU_BOTTLENECK_LAYERS']
# Downsampling path and recover skip connections for each transition down block
x = Conv2D(init_filters, (3, 3),
kernel_initializer=params['WEIGHTS_INIT'],
padding='same',
name='conv_initial')(image)
prev_filters = init_filters
skip_conn = []
for td in range(num_transition_blocks):
nb_filters_conv = prev_filters + n_layers_down[td] * growth_rate
[x, skip] = self.add_transitiondown_block(x, nb_filters_conv, 2,
params['WEIGHTS_INIT'],
n_layers_down[td],
growth_rate,
params['DROPOUT_P'])
skip_conn.append(skip)
prev_filters = nb_filters_conv
# Middle of the path (bottleneck)
x = self.add_dense_block(
x, bottleneck_layers, growth_rate, params['DROPOUT_P'],
params['WEIGHTS_INIT']) # feature maps: 512 input, 592 output
# Upsampling path
skip_conn = skip_conn[::-1]
prev_filters = bottleneck_layers * growth_rate
for tu in range(num_transition_blocks):
x = self.add_transitionup_block(x, skip_conn[tu], prev_filters,
params['WEIGHTS_INIT'],
n_layers_up[tu], growth_rate,
params['DROPOUT_P'])
prev_filters = n_layers_up[tu] * growth_rate
# Final classification layer (batch_size, classes, width, height)
x = Conv2D(params['NUM_CLASSES'], (1, 1),
kernel_initializer=params['WEIGHTS_INIT'],
padding='same')(x)
# Reshape to (None, width*height, classes) before applying softmax
x = Lambda(
lambda x: x.flatten(ndim=3),
output_shape=lambda s: tuple([s[0], params['NUM_CLASSES'], None]))(
x)
x = Permute((2, 1))(x)
matrix_out = Activation(params['CLASSIFIER_ACTIVATION'],
name=self.ids_outputs[0])(x)
self.model = Model(inputs=image, outputs=matrix_out)
def ClassicUpsampling(self, params):
"""
References:
Olaf Ronneberger et al. U-net: Convolutional networks for biomedical image segmentation.
"""
self.ids_inputs = params["INPUTS_IDS_MODEL"]
self.ids_outputs = params["OUTPUTS_IDS_MODEL"]
crop = params['IMG_CROP_SIZE']
image = Input(shape=tuple([crop[-1], None, None]),
name=self.ids_inputs[0])
# Downsampling path and recover skip connections for each transition down block
concat_axis = 1
conv1 = Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='conv1_1')(image)
conv1 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv1)
conv1 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv1)
pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
conv2 = Conv2D(128, (3, 3), activation='relu', padding='same')(pool1)
conv2 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv2)
conv2 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv2)
pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(pool2)
conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv3)
conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv3)
pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
conv4 = Conv2D(256, (3, 3), activation='relu', padding='same')(pool3)
conv4 = Conv2D(256, (3, 3), activation='relu', padding='same')(conv4)
conv4 = Conv2D(256, (3, 3), activation='relu', padding='same')(conv4)
pool4 = MaxPooling2D(pool_size=(2, 2))(conv4)
# Middle of the path (bottleneck)
conv5 = Conv2D(512, (3, 3), activation='relu', padding='same')(pool4)
conv5 = Conv2D(512, (3, 3), activation='relu', padding='same')(conv5)
conv5 = Conv2D(512, (3, 3), activation='relu', padding='same')(conv5)
# Upsampling path
up_conv5 = UpSampling2D(size=(2, 2))(conv5)
up_conv5 = ZeroPadding2D()(up_conv5)
up6 = Concat(cropping=[None, None, 'center', 'center'])(
[conv4, up_conv5])
conv6 = Conv2D(256, (3, 3), activation='relu', padding='same')(up6)
conv6 = Conv2D(256, (3, 3), activation='relu', padding='same')(conv6)
conv6 = Conv2D(256, (3, 3), activation='relu', padding='same')(conv6)
up_conv6 = UpSampling2D(size=(2, 2))(conv6)
up_conv6 = ZeroPadding2D()(up_conv6)
up7 = Concat(cropping=[None, None, 'center', 'center'])(
[conv3, up_conv6])
conv7 = Conv2D(128, (3, 3), activation='relu', padding='same')(up7)
conv7 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv7)
conv7 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv7)
up_conv7 = UpSampling2D(size=(2, 2))(conv7)
up_conv7 = ZeroPadding2D()(up_conv7)
up8 = Concat(cropping=[None, None, 'center', 'center'])(
[conv2, up_conv7])
conv8 = Conv2D(128, (3, 3), activation='relu', padding='same')(up8)
conv8 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv8)
conv8 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv8)
up_conv8 = UpSampling2D(size=(2, 2))(conv8)
up_conv8 = ZeroPadding2D()(up_conv8)
up9 = Concat(cropping=[None, None, 'center', 'center'])(
[conv1, up_conv8])
conv9 = Conv2D(64, (3, 3), activation='relu', padding='same')(up9)
conv9 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv9)
conv9 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv9)
# Final classification layer (batch_size, classes, width, height)
x = Conv2D(params['NUM_CLASSES'], (1, 1), border_mode='same')(conv9)
# Reshape to (None, width*height, classes) before applying softmax
x = Lambda(
lambda x: x.flatten(ndim=3),
output_shape=lambda s: tuple([s[0], params['NUM_CLASSES'], None]))(
x)
x = Permute((2, 1))(x)
matrix_out = Activation(params['CLASSIFIER_ACTIVATION'],
name=self.ids_outputs[0])(x)
self.model = Model(inputs=image, outputs=matrix_out)