def __init__(self, class_no=14):
print('create object _unet')
self.class_no = class_no
self.kernel_size1 = 1
self.kernel_size2 = 3
self.log_ext = '_'
self.seed = 200
self.upsampling3d = upsampling()
self.layers = layers.layers()
def __init__(self, class_no=14):
print('create object _unet')
self.class_no = class_no
self.kernel_size1 = 1
self.kernel_size2 = 3
self.log_ext = '_'
self.seed_no=200
self.upsampling3d=upsampling()
self.maxpool = False
self.norm_method = 'batch_normalization'
self.layers=layers()
def __init__(self, graph, class_no=14):
print('create object _unet')
self.graph = graph
self.class_no = class_no
self.kernel_size1 = 1
self.kernel_size2 = 3
self.log_ext = '_'
self.seed = 200
self.upsampling3d = upsampling()
self.layers = layers()
self.reuse = False
self.trainable = True
def __init__(self, trainable, file_name):
print('create object _unet')
self.upsampling3d = upsampling()
self.layers = layers()
self.trainable = trainable
self.seed = 200
self.kernel_loader = loader(file_name)
# self.kernel_loader.print_tensors_in_checkpoint_file(file_name, False,
# True, False)
[
self.conv_init1_ld1, self.bias_init1_ld1, self.beta_init1_ld1,
self.gamma_init1_ld1, self.moving_mean_init1_ld1,
self.moving_var1_ld1, self.conv_init2_ld1, self.bias_init2_ld1,
self.beta_init2_ld1, self.gamma_init2_ld1,
self.moving_mean_init2_ld1, self.moving_var2_ld1,
self.conv_init1_ld2, self.bias_init1_ld2, self.beta_init1_ld2,
self.gamma_init1_ld2, self.moving_mean_init1_ld2,
self.moving_var1_ld2, self.conv_init2_ld2, self.bias_init2_ld2,
self.beta_init2_ld2, self.gamma_init2_ld2,
self.moving_mean_init2_ld2, self.moving_var2_ld2,
self.conv_init1_ld3, self.bias_init1_ld3, self.beta_init1_ld3,
self.gamma_init1_ld3, self.moving_mean_init1_ld3,
self.moving_var1_ld3, self.conv_init2_ld3, self.bias_init2_ld3,
self.beta_init2_ld3, self.gamma_init2_ld3,
self.moving_mean_init2_ld3, self.moving_var2_ld3
] = self.kernel_loader.return_tensor_value_list_by_name([
'U_LD_DS1/U_LD_DS1U_conv1_conv3d/kernel',
'U_LD_DS1/U_LD_DS1U_conv1_conv3d/bias',
'U_LD_DS1/U_LD_DS1U_conv1_bn/beta',
'U_LD_DS1/U_LD_DS1U_conv1_bn/gamma',
'U_LD_DS1/U_LD_DS1U_conv1_bn/moving_mean',
'U_LD_DS1/U_LD_DS1U_conv1_bn/moving_variance',
'U_LD_DS1/U_LD_DS1U_conv2_conv3d/kernel',
'U_LD_DS1/U_LD_DS1U_conv2_conv3d/bias',
'U_LD_DS1/U_LD_DS1U_conv2_bn/beta',
'U_LD_DS1/U_LD_DS1U_conv2_bn/gamma',
'U_LD_DS1/U_LD_DS1U_conv2_bn/moving_mean',
'U_LD_DS1/U_LD_DS1U_conv2_bn/moving_variance',
'U_LD_DS2/U_LD_DS2U_conv1_conv3d/kernel',
'U_LD_DS2/U_LD_DS2U_conv1_conv3d/bias',
'U_LD_DS2/U_LD_DS2U_conv1_bn/beta',
'U_LD_DS2/U_LD_DS2U_conv1_bn/gamma',
'U_LD_DS2/U_LD_DS2U_conv1_bn/moving_mean',
'U_LD_DS2/U_LD_DS2U_conv1_bn/moving_variance',
'U_LD_DS2/U_LD_DS2U_conv2_conv3d/kernel',
'U_LD_DS2/U_LD_DS2U_conv2_conv3d/bias',
'U_LD_DS2/U_LD_DS2U_conv2_bn/beta',
'U_LD_DS2/U_LD_DS2U_conv2_bn/gamma',
'U_LD_DS2/U_LD_DS2U_conv2_bn/moving_mean',
'U_LD_DS2/U_LD_DS2U_conv2_bn/moving_variance',
'U_LD_US1/U_LD_US1U_conv1_conv3d/kernel',
'U_LD_US1/U_LD_US1U_conv1_conv3d/bias',
'U_LD_US1/U_LD_US1U_conv1_bn/beta',
'U_LD_US1/U_LD_US1U_conv1_bn/gamma',
'U_LD_US1/U_LD_US1U_conv1_bn/moving_mean',
'U_LD_US1/U_LD_US1U_conv1_bn/moving_variance',
'U_LD_US1/U_LD_US1U_conv2_conv3d/kernel',
'U_LD_US1/U_LD_US1U_conv2_conv3d/bias',
'U_LD_US1/U_LD_US1U_conv2_bn/beta',
'U_LD_US1/U_LD_US1U_conv2_bn/gamma',
'U_LD_US1/U_LD_US1U_conv2_bn/moving_mean',
'U_LD_US1/U_LD_US1U_conv2_bn/moving_variance',
])
unet = _unet(trainable=False)
if save_pb:
if not os.path.exists(path + '/pbs'):
os.makedirs(path + '/pbs')
freeze_graph(half_unet_graph_chckpnt_dir,
path + '/pbs/{}.pb'.format('jpg'), 'U_y/U_y_conv3d/bias')
#========================
# show all nodes of a graph
AA = [n.name for n in tf.get_default_graph().as_graph_def().node]
for i in AA:
print(i)
#========================
layers = layers()
X = tf.placeholder(tf.float32,
shape=[None, None, None, None, 1],
name='synth_img_row1')
conv1 = layers.conv3d(input,
filters=10,
kernel_size=3,
padding='same',
dilation_rate=1,
is_training=True,
trainable='True',
scope='conv1',
reuse='False')
unet.unet(conv1)
#
#