deepreg.model.network.DVFModel.build_loss() - Code Metrics - Inspection of "Merge pull request #713 from DeepRegNet/708-log-mo..." - DeepRegNet/DeepReg - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Push — main ( 3b5a47...3ab0c2 )

unknown

created 2021-03-26 22:57 UTC

deepreg.model.network.DVFModel.build_loss() A

↳ Parent: deepreg.model.network

Complexity

Conditions

Size

Total Lines	7
Code Lines	4

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
eloc	4
dl	0
loc	7
rs	10
c	0
b	0
f	0
cc	1
nop	1

import logging
import os
from abc import abstractmethod
from copy import deepcopy
from typing import Dict, Optional, Tuple

import tensorflow as tf

from deepreg.loss.label import DiceScore, compute_centroid_distance
from deepreg.model import layer, layer_util
from deepreg.model.backbone import GlobalNet
from deepreg.registry import REGISTRY


def dict_without(d: dict, key) -> dict:
    """
    Return a copy of the given dict without a certain key.

    :param d: dict to be copied.
    :param key: key to be removed.
    :return: the copy without a key
    """
    copied = deepcopy(d)
    copied.pop(key)
    return copied


class RegistrationModel(tf.keras.Model):
    """Interface for registration model."""

    def __init__(
        self,
        moving_image_size: tuple,
        fixed_image_size: tuple,
        index_size: int,
        labeled: bool,
        batch_size: int,
        config: dict,
        num_devices: int = 1,
        name: str = "RegistrationModel",
    ):
        """
        Init.

        :param moving_image_size: (m_dim1, m_dim2, m_dim3)
        :param fixed_image_size: (f_dim1, f_dim2, f_dim3)
        :param index_size: number of indices for identify each sample
        :param labeled: if the data is labeled
        :param batch_size: size of mini-batch
        :param config: config for method, backbone, and loss.
        :param num_devices: number of GPU used,
            global_batch_size = batch_size*num_devices
        :param name: name of the model
        """
        super().__init__(name=name)
        self.moving_image_size = moving_image_size
        self.fixed_image_size = fixed_image_size
        self.index_size = index_size
        self.labeled = labeled
        self.batch_size = batch_size
        self.config = config
        self.num_devices = num_devices
        self.global_batch_size = num_devices * batch_size

        self._inputs = None  # save inputs of self._model as dict
        self._outputs = None  # save outputs of self._model as dict

        self.grid_ref = layer_util.get_reference_grid(grid_size=fixed_image_size)[
            None, ...
        ]
        self._model: tf.keras.Model = self.build_model()
        self.build_loss()

    def get_config(self) -> dict:
        """Return the config dictionary for recreating this class."""
        return dict(
            moving_image_size=self.moving_image_size,
            fixed_image_size=self.fixed_image_size,
            index_size=self.index_size,
            labeled=self.labeled,
            batch_size=self.batch_size,
            config=self.config,
            num_devices=self.num_devices,
            name=self.name,
        )

    @abstractmethod
    def build_model(self):
        """Build the model to be saved as self._model."""

    def build_inputs(self) -> Dict[str, tf.keras.layers.Input]:
        """
        Build input tensors.

        :return: dict of inputs.
        """
        # (batch, m_dim1, m_dim2, m_dim3, 1)
        moving_image = tf.keras.Input(
            shape=self.moving_image_size,
            batch_size=self.batch_size,
            name="moving_image",
        )
        # (batch, f_dim1, f_dim2, f_dim3, 1)
        fixed_image = tf.keras.Input(
            shape=self.fixed_image_size,
            batch_size=self.batch_size,
            name="fixed_image",
        )
        # (batch, index_size)
        indices = tf.keras.Input(
            shape=(self.index_size,),
            batch_size=self.batch_size,
            name="indices",
        )

        if not self.labeled:
            return dict(
                moving_image=moving_image, fixed_image=fixed_image, indices=indices
            )

        # (batch, m_dim1, m_dim2, m_dim3, 1)
        moving_label = tf.keras.Input(
            shape=self.moving_image_size,
            batch_size=self.batch_size,
            name="moving_label",
        )
        # (batch, m_dim1, m_dim2, m_dim3, 1)
        fixed_label = tf.keras.Input(
            shape=self.fixed_image_size,
            batch_size=self.batch_size,
            name="fixed_label",
        )
        return dict(
            moving_image=moving_image,
            fixed_image=fixed_image,
            moving_label=moving_label,
            fixed_label=fixed_label,
            indices=indices,
        )

    def concat_images(
        self,
        moving_image: tf.Tensor,
        fixed_image: tf.Tensor,
        moving_label: Optional[tf.Tensor] = None,
    ) -> tf.Tensor:
        """
        Adjust image shape and concatenate them together.

        :param moving_image: registration source
        :param fixed_image: registration target
        :param moving_label: optional, only used for conditional model.
        :return:
        """
        images = []

        resize_layer = layer.Resize3d(shape=self.fixed_image_size)

        # (batch, m_dim1, m_dim2, m_dim3, 1)
        moving_image = tf.expand_dims(moving_image, axis=4)
        moving_image = resize_layer(moving_image)
        images.append(moving_image)

        # (batch, m_dim1, m_dim2, m_dim3, 1)
        fixed_image = tf.expand_dims(fixed_image, axis=4)
        images.append(fixed_image)

        # (batch, m_dim1, m_dim2, m_dim3, 1)
        if moving_label is not None:
            moving_label = tf.expand_dims(moving_label, axis=4)
            moving_label = resize_layer(moving_label)
            images.append(moving_label)

        # (batch, f_dim1, f_dim2, f_dim3, 2 or 3)
        images = tf.concat(images, axis=4)
        return images

    def _build_loss(self, name: str, inputs_dict: dict):
        """
        Build and add one weighted loss together with the metrics.

        :param name: name of loss, image / label / regularization.
        :param inputs_dict: inputs for loss function
        """

        if name not in self.config["loss"]:
            # loss config is not defined
            logging.warning(
                f"The configuration for loss {name} is not defined. "
                f"Therefore it is not used."
            )
            return

        loss_configs = self.config["loss"][name]
        if not isinstance(loss_configs, list):
            loss_configs = [loss_configs]

        for loss_config in loss_configs:

            if "weight" not in loss_config:
                # default loss weight 1
                logging.warning(
                    f"The weight for loss {name} is not defined."
                    f"Default weight = 1.0 is used."
                )
                loss_config["weight"] = 1.0

            # build loss
            weight = loss_config["weight"]

            if weight == 0:
                logging.warning(
                    f"The weight for loss {name} is zero." f"Loss is not used."
                )
                return

            loss_layer: tf.keras.layers.Layer = REGISTRY.build_loss(
                config=dict_without(d=loss_config, key="weight")
            )
            loss_value = loss_layer(**inputs_dict) / self.global_batch_size
            weighted_loss = loss_value * weight

            # add loss
            self._model.add_loss(weighted_loss)

            # add metric
            self._model.add_metric(
                loss_value, name=f"loss/{name}_{loss_layer.name}", aggregation="mean"
            )
            self._model.add_metric(
                weighted_loss,
                name=f"loss/{name}_{loss_layer.name}_weighted",
                aggregation="mean",
            )

    @abstractmethod
    def build_loss(self):
        """Build losses according to configs."""

        # input metrics
        fixed_image = self._inputs["fixed_image"]
        moving_image = self._inputs["moving_image"]
        self.log_tensor_stats(tensor=moving_image, name="moving_image")
        self.log_tensor_stats(tensor=fixed_image, name="fixed_image")

        # image loss, conditional model does not have this
        if "pred_fixed_image" in self._outputs:
            pred_fixed_image = self._outputs["pred_fixed_image"]
            self._build_loss(
                name="image",
                inputs_dict=dict(y_true=fixed_image, y_pred=pred_fixed_image),
            )

        if self.labeled:
            # input metrics
            fixed_label = self._inputs["fixed_label"]
            moving_label = self._inputs["moving_label"]
            self.log_tensor_stats(tensor=moving_label, name="moving_label")
            self.log_tensor_stats(tensor=fixed_label, name="fixed_label")

            # label loss
            pred_fixed_label = self._outputs["pred_fixed_label"]
            self._build_loss(
                name="label",
                inputs_dict=dict(y_true=fixed_label, y_pred=pred_fixed_label),
            )

            # additional label metrics
            tre = compute_centroid_distance(
                y_true=fixed_label, y_pred=pred_fixed_label, grid=self.grid_ref
            )
            dice_binary = DiceScore(binary=True)(
                y_true=fixed_label, y_pred=pred_fixed_label
            )
            self._model.add_metric(tre, name="metric/TRE", aggregation="mean")
            self._model.add_metric(
                dice_binary, name="metric/BinaryDiceScore", aggregation="mean"
            )

    def call(
        self, inputs: Dict[str, tf.Tensor], training=None, mask=None
    ) -> Dict[str, tf.Tensor]:
        """
        Call the self._model.

        :param inputs: a dict of tensors.
        :param training: training or not.
        :param mask: maks for inputs.
        :return:
        """
        return self._model(inputs, training=training, mask=mask)  # pragma: no cover

    @abstractmethod
    def postprocess(
        self,
        inputs: Dict[str, tf.Tensor],
        outputs: Dict[str, tf.Tensor],
    ) -> Tuple[tf.Tensor, Dict]:
        """
        Return a dict used for saving inputs and outputs.

        :param inputs: dict of model inputs
        :param outputs: dict of model outputs
        :return: tuple, indices and a dict.
            In the dict, each value is (tensor, normalize, on_label), where
            - normalize = True if the tensor need to be normalized to [0, 1]
            - on_label = True if the tensor depends on label
        """

    def plot_model(self, output_dir: str):
        """
        Save model structure in png.

        :param output_dir: path to the output dir.
        """
        logging.info(self._model.summary())
        try:
            tf.keras.utils.plot_model(
                self._model,
                to_file=os.path.join(output_dir, f"{self.name}.png"),
                dpi=96,
                show_shapes=True,
                show_layer_names=True,
                expand_nested=False,
            )
        except ImportError as err:  # pragma: no cover
            logging.error(
                "Failed to plot model structure."
                "Please check if graphviz is installed.\n"
                "Error message is:"
                f"{err}"
            )

    def log_tensor_stats(self, tensor: tf.Tensor, name: str):
        """
        Log statistics of a given tensor.

        :param tensor: tensor to monitor.
        :param name: name of the tensor.
        """
        flatten = tf.reshape(tensor, shape=(self.batch_size, -1))
        self._model.add_metric(
            tf.reduce_mean(flatten, axis=1),
            name=f"metric/{name}_mean",
            aggregation="mean",
        )
        self._model.add_metric(
            tf.reduce_min(flatten, axis=1),
            name=f"metric/{name}_min",
            aggregation="min",
        )
        self._model.add_metric(
            tf.reduce_max(flatten, axis=1),
            name=f"metric/{name}_max",
            aggregation="max",
        )


@REGISTRY.register_model(name="ddf")
class DDFModel(RegistrationModel):
    """
    A registration model predicts DDF.

    When using global net as backbone,
    the model predicts an affine transformation parameters,
    and a DDF is calculated based on that.
    """

    name = "DDFModel"

    def _resize_interpolate(self, field, control_points):
        resize = layer.ResizeCPTransform(control_points)
        field = resize(field)

        interpolate = layer.BSplines3DTransform(control_points, self.fixed_image_size)
        field = interpolate(field)

        return field

    def build_model(self):
        """Build the model to be saved as self._model."""
        # build inputs
        self._inputs = self.build_inputs()
        moving_image = self._inputs["moving_image"]
        fixed_image = self._inputs["fixed_image"]

        # build ddf
        control_points = self.config["backbone"].pop("control_points", False)
        backbone_inputs = self.concat_images(moving_image, fixed_image)
        backbone = REGISTRY.build_backbone(
            config=self.config["backbone"],
            default_args=dict(
                image_size=self.fixed_image_size,
                out_channels=3,
                out_kernel_initializer="zeros",
                out_activation=None,
            ),
        )

        if isinstance(backbone, GlobalNet):
            # (f_dim1, f_dim2, f_dim3, 3), (4, 3)
            ddf, theta = backbone(inputs=backbone_inputs)
            self._outputs = dict(ddf=ddf, theta=theta)
        else:
            # (f_dim1, f_dim2, f_dim3, 3)
            ddf = backbone(inputs=backbone_inputs)
            ddf = (
                self._resize_interpolate(ddf, control_points) if control_points else ddf
            )
            self._outputs = dict(ddf=ddf)

        # build outputs
        warping = layer.Warping(fixed_image_size=self.fixed_image_size)
        # (f_dim1, f_dim2, f_dim3, 3)
        pred_fixed_image = warping(inputs=[ddf, moving_image])
        self._outputs["pred_fixed_image"] = pred_fixed_image

        if not self.labeled:
            return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)

        # (f_dim1, f_dim2, f_dim3, 3)
        moving_label = self._inputs["moving_label"]
        pred_fixed_label = warping(inputs=[ddf, moving_label])

        self._outputs["pred_fixed_label"] = pred_fixed_label
        return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)

    def build_loss(self):
        """Build losses according to configs."""
        super().build_loss()

        # ddf loss and metrics
        ddf = self._outputs["ddf"]
        self._build_loss(name="regularization", inputs_dict=dict(inputs=ddf))
        self.log_tensor_stats(tensor=ddf, name="ddf")

    def postprocess(
        self,
        inputs: Dict[str, tf.Tensor],
        outputs: Dict[str, tf.Tensor],
    ) -> Tuple[tf.Tensor, Dict]:
        """
        Return a dict used for saving inputs and outputs.

        :param inputs: dict of model inputs
        :param outputs: dict of model outputs
        :return: tuple, indices and a dict.
            In the dict, each value is (tensor, normalize, on_label), where
            - normalize = True if the tensor need to be normalized to [0, 1]
            - on_label = True if the tensor depends on label
        """
        indices = inputs["indices"]
        processed = dict(
            moving_image=(inputs["moving_image"], True, False),
            fixed_image=(inputs["fixed_image"], True, False),
            ddf=(outputs["ddf"], True, False),
            pred_fixed_image=(outputs["pred_fixed_image"], True, False),
        )

        # save theta for affine model
        if "theta" in outputs:
            processed["theta"] = (outputs["theta"], None, None)  # type: ignore

        if not self.labeled:
            return indices, processed

        processed = {
            **dict(
                moving_label=(inputs["moving_label"], False, True),
                fixed_label=(inputs["fixed_label"], False, True),
                pred_fixed_label=(outputs["pred_fixed_label"], False, True),
            ),
            **processed,
        }

        return indices, processed


@REGISTRY.register_model(name="dvf")
class DVFModel(DDFModel):
    """
    A registration model predicts DVF.

    DDF is calculated based on DVF.
    """

    name = "DVFModel"

    def build_model(self):
        """Build the model to be saved as self._model."""
        # build inputs
        self._inputs = self.build_inputs()
        moving_image = self._inputs["moving_image"]
        fixed_image = self._inputs["fixed_image"]
        control_points = self.config["backbone"].pop("control_points", False)

        # build ddf
        backbone_inputs = self.concat_images(moving_image, fixed_image)
        backbone = REGISTRY.build_backbone(
            config=self.config["backbone"],
            default_args=dict(
                image_size=self.fixed_image_size,
                out_channels=3,
                out_kernel_initializer="zeros",
                out_activation=None,
            ),
        )
        dvf = backbone(inputs=backbone_inputs)
        dvf = self._resize_interpolate(dvf, control_points) if control_points else dvf
        ddf = layer.IntDVF(fixed_image_size=self.fixed_image_size)(dvf)

        # build outputs
        self._warping = layer.Warping(fixed_image_size=self.fixed_image_size)
        # (f_dim1, f_dim2, f_dim3, 3)
        pred_fixed_image = self._warping(inputs=[ddf, moving_image])

        self._outputs = dict(dvf=dvf, ddf=ddf, pred_fixed_image=pred_fixed_image)

        if not self.labeled:
            return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)

        # (f_dim1, f_dim2, f_dim3, 3)
        moving_label = self._inputs["moving_label"]
        pred_fixed_label = self._warping(inputs=[ddf, moving_label])

        self._outputs["pred_fixed_label"] = pred_fixed_label
        return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)

    def build_loss(self):
        """Build losses according to configs."""
        super().build_loss()

        # dvf metrics
        dvf = self._outputs["dvf"]
        self.log_tensor_stats(tensor=dvf, name="dvf")

    def postprocess(
        self,
        inputs: Dict[str, tf.Tensor],
        outputs: Dict[str, tf.Tensor],
    ) -> Tuple[tf.Tensor, Dict]:
        """
        Return a dict used for saving inputs and outputs.

        :param inputs: dict of model inputs
        :param outputs: dict of model outputs
        :return: tuple, indices and a dict.
            In the dict, each value is (tensor, normalize, on_label), where
            - normalize = True if the tensor need to be normalized to [0, 1]
            - on_label = True if the tensor depends on label
        """
        indices, processed = super().postprocess(inputs=inputs, outputs=outputs)
        processed["dvf"] = (outputs["dvf"], True, False)
        return indices, processed


@REGISTRY.register_model(name="conditional")
class ConditionalModel(RegistrationModel):
    """
    A registration model predicts fixed image label without DDF or DVF.
    """

    name = "ConditionalModel"

    def build_model(self):
        """Build the model to be saved as self._model."""
        assert self.labeled

        # build inputs
        self._inputs = self.build_inputs()
        moving_image = self._inputs["moving_image"]
        fixed_image = self._inputs["fixed_image"]
        moving_label = self._inputs["moving_label"]

        # build ddf
        backbone_inputs = self.concat_images(moving_image, fixed_image, moving_label)
        backbone = REGISTRY.build_backbone(
            config=self.config["backbone"],
            default_args=dict(
                image_size=self.fixed_image_size,
                out_channels=1,
                out_kernel_initializer="glorot_uniform",
                out_activation="sigmoid",
            ),
        )
        # (batch, f_dim1, f_dim2, f_dim3)
        pred_fixed_label = backbone(inputs=backbone_inputs)
        pred_fixed_label = tf.squeeze(pred_fixed_label, axis=4)

        self._outputs = dict(pred_fixed_label=pred_fixed_label)
        return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)

    def postprocess(
        self,
        inputs: Dict[str, tf.Tensor],
        outputs: Dict[str, tf.Tensor],
    ) -> Tuple[tf.Tensor, Dict]:
        """
        Return a dict used for saving inputs and outputs.

        :param inputs: dict of model inputs
        :param outputs: dict of model outputs
        :return: tuple, indices and a dict.
            In the dict, each value is (tensor, normalize, on_label), where
            - normalize = True if the tensor need to be normalized to [0, 1]
            - on_label = True if the tensor depends on label
        """
        indices = inputs["indices"]
        processed = dict(
            moving_image=(inputs["moving_image"], True, False),
            fixed_image=(inputs["fixed_image"], True, False),
            pred_fixed_label=(outputs["pred_fixed_label"], True, True),
            moving_label=(inputs["moving_label"], False, True),
            fixed_label=(inputs["fixed_label"], False, True),
        )

        return indices, processed


1			import logging
2			import os
3			from abc import abstractmethod
4			from copy import deepcopy
5			from typing import Dict, Optional, Tuple
6
7			import tensorflow as tf
8
9			from deepreg.loss.label import DiceScore, compute_centroid_distance
10			from deepreg.model import layer, layer_util
11			from deepreg.model.backbone import GlobalNet
12			from deepreg.registry import REGISTRY
13
14
15			def dict_without(d: dict, key) -> dict:
16			"""
17			Return a copy of the given dict without a certain key.
18
19			:param d: dict to be copied.
20			:param key: key to be removed.
21			:return: the copy without a key
22			"""
23			copied = deepcopy(d)
24			copied.pop(key)
25			return copied
26
27
28			class RegistrationModel(tf.keras.Model):
29			"""Interface for registration model."""
30
31			def __init__(
32			self,
33			moving_image_size: tuple,
34			fixed_image_size: tuple,
35			index_size: int,
36			labeled: bool,
37			batch_size: int,
38			config: dict,
39			num_devices: int = 1,
40			name: str = "RegistrationModel",
41			):
42			"""
43			Init.
44
45			:param moving_image_size: (m_dim1, m_dim2, m_dim3)
46			:param fixed_image_size: (f_dim1, f_dim2, f_dim3)
47			:param index_size: number of indices for identify each sample
48			:param labeled: if the data is labeled
49			:param batch_size: size of mini-batch
50			:param config: config for method, backbone, and loss.
51			:param num_devices: number of GPU used,
52			global_batch_size = batch_size*num_devices
53			:param name: name of the model
54			"""
55			super().__init__(name=name)
56			self.moving_image_size = moving_image_size
57			self.fixed_image_size = fixed_image_size
58			self.index_size = index_size
59			self.labeled = labeled
60			self.batch_size = batch_size
61			self.config = config
62			self.num_devices = num_devices
63			self.global_batch_size = num_devices * batch_size
64
65			self._inputs = None # save inputs of self._model as dict
66			self._outputs = None # save outputs of self._model as dict
67
68			self.grid_ref = layer_util.get_reference_grid(grid_size=fixed_image_size)[
69			None, ...
70			]
71			self._model: tf.keras.Model = self.build_model()
72			self.build_loss()
73
74			def get_config(self) -> dict:
75			"""Return the config dictionary for recreating this class."""
76			return dict(
77			moving_image_size=self.moving_image_size,
78			fixed_image_size=self.fixed_image_size,
79			index_size=self.index_size,
80			labeled=self.labeled,
81			batch_size=self.batch_size,
82			config=self.config,
83			num_devices=self.num_devices,
84			name=self.name,
85			)
86
87			@abstractmethod
88			def build_model(self):
89			"""Build the model to be saved as self._model."""
90
91			def build_inputs(self) -> Dict[str, tf.keras.layers.Input]:
92			"""
93			Build input tensors.
94
95			:return: dict of inputs.
96			"""
97			# (batch, m_dim1, m_dim2, m_dim3, 1)
98			moving_image = tf.keras.Input(
99			shape=self.moving_image_size,
100			batch_size=self.batch_size,
101			name="moving_image",
102			)
103			# (batch, f_dim1, f_dim2, f_dim3, 1)
104			fixed_image = tf.keras.Input(
105			shape=self.fixed_image_size,
106			batch_size=self.batch_size,
107			name="fixed_image",
108			)
109			# (batch, index_size)
110			indices = tf.keras.Input(
111			shape=(self.index_size,),
112			batch_size=self.batch_size,
113			name="indices",
114			)
115
116			if not self.labeled:
117			return dict(
118			moving_image=moving_image, fixed_image=fixed_image, indices=indices
119			)
120
121			# (batch, m_dim1, m_dim2, m_dim3, 1)
122			moving_label = tf.keras.Input(
123			shape=self.moving_image_size,
124			batch_size=self.batch_size,
125			name="moving_label",
126			)
127			# (batch, m_dim1, m_dim2, m_dim3, 1)
128			fixed_label = tf.keras.Input(
129			shape=self.fixed_image_size,
130			batch_size=self.batch_size,
131			name="fixed_label",
132			)
133			return dict(
134			moving_image=moving_image,
135			fixed_image=fixed_image,
136			moving_label=moving_label,
137			fixed_label=fixed_label,
138			indices=indices,
139			)
140
141			def concat_images(
142			self,
143			moving_image: tf.Tensor,
144			fixed_image: tf.Tensor,
145			moving_label: Optional[tf.Tensor] = None,
146			) -> tf.Tensor:
147			"""
148			Adjust image shape and concatenate them together.
149
150			:param moving_image: registration source
151			:param fixed_image: registration target
152			:param moving_label: optional, only used for conditional model.
153			:return:
154			"""
155			images = []
156
157			resize_layer = layer.Resize3d(shape=self.fixed_image_size)
158
159			# (batch, m_dim1, m_dim2, m_dim3, 1)
160			moving_image = tf.expand_dims(moving_image, axis=4)
161			moving_image = resize_layer(moving_image)
162			images.append(moving_image)
163
164			# (batch, m_dim1, m_dim2, m_dim3, 1)
165			fixed_image = tf.expand_dims(fixed_image, axis=4)
166			images.append(fixed_image)
167
168			# (batch, m_dim1, m_dim2, m_dim3, 1)
169			if moving_label is not None:
170			moving_label = tf.expand_dims(moving_label, axis=4)
171			moving_label = resize_layer(moving_label)
172			images.append(moving_label)
173
174			# (batch, f_dim1, f_dim2, f_dim3, 2 or 3)
175			images = tf.concat(images, axis=4)
176			return images
177
178			def _build_loss(self, name: str, inputs_dict: dict):
179			"""
180			Build and add one weighted loss together with the metrics.
181
182			:param name: name of loss, image / label / regularization.
183			:param inputs_dict: inputs for loss function
184			"""
185
186			if name not in self.config["loss"]:
187			# loss config is not defined
188			logging.warning(
189			f"The configuration for loss {name} is not defined. "
190			f"Therefore it is not used."
191			)
192			return
193
194			loss_configs = self.config["loss"][name]
195			if not isinstance(loss_configs, list):
196			loss_configs = [loss_configs]
197
198			for loss_config in loss_configs:
199
200			if "weight" not in loss_config:
201			# default loss weight 1
202			logging.warning(
203			f"The weight for loss {name} is not defined."
204			f"Default weight = 1.0 is used."
205			)
206			loss_config["weight"] = 1.0
207
208			# build loss
209			weight = loss_config["weight"]
210
211			if weight == 0:
212			logging.warning(
213			f"The weight for loss {name} is zero." f"Loss is not used."
214			)
215			return
216
217			loss_layer: tf.keras.layers.Layer = REGISTRY.build_loss(
218			config=dict_without(d=loss_config, key="weight")
219			)
220			loss_value = loss_layer(**inputs_dict) / self.global_batch_size
221			weighted_loss = loss_value * weight
222
223			# add loss
224			self._model.add_loss(weighted_loss)
225
226			# add metric
227			self._model.add_metric(
228			loss_value, name=f"loss/{name}_{loss_layer.name}", aggregation="mean"
229			)
230			self._model.add_metric(
231			weighted_loss,
232			name=f"loss/{name}_{loss_layer.name}_weighted",
233			aggregation="mean",
234			)
235
236			@abstractmethod
237			def build_loss(self):
238			"""Build losses according to configs."""
239
240			# input metrics
241			fixed_image = self._inputs["fixed_image"]
242			moving_image = self._inputs["moving_image"]
243			self.log_tensor_stats(tensor=moving_image, name="moving_image")
244			self.log_tensor_stats(tensor=fixed_image, name="fixed_image")
245
246			# image loss, conditional model does not have this
247			if "pred_fixed_image" in self._outputs:
248			pred_fixed_image = self._outputs["pred_fixed_image"]
249			self._build_loss(
250			name="image",
251			inputs_dict=dict(y_true=fixed_image, y_pred=pred_fixed_image),
252			)
253
254			if self.labeled:
255			# input metrics
256			fixed_label = self._inputs["fixed_label"]
257			moving_label = self._inputs["moving_label"]
258			self.log_tensor_stats(tensor=moving_label, name="moving_label")
259			self.log_tensor_stats(tensor=fixed_label, name="fixed_label")
260
261			# label loss
262			pred_fixed_label = self._outputs["pred_fixed_label"]
263			self._build_loss(
264			name="label",
265			inputs_dict=dict(y_true=fixed_label, y_pred=pred_fixed_label),
266			)
267
268			# additional label metrics
269			tre = compute_centroid_distance(
270			y_true=fixed_label, y_pred=pred_fixed_label, grid=self.grid_ref
271			)
272			dice_binary = DiceScore(binary=True)(
273			y_true=fixed_label, y_pred=pred_fixed_label
274			)
275			self._model.add_metric(tre, name="metric/TRE", aggregation="mean")
276			self._model.add_metric(
277			dice_binary, name="metric/BinaryDiceScore", aggregation="mean"
278			)
279
280			def call(
281			self, inputs: Dict[str, tf.Tensor], training=None, mask=None
282			) -> Dict[str, tf.Tensor]:
283			"""
284			Call the self._model.
285
286			:param inputs: a dict of tensors.
287			:param training: training or not.
288			:param mask: maks for inputs.
289			:return:
290			"""
291			return self._model(inputs, training=training, mask=mask) # pragma: no cover
292
293			@abstractmethod
294			def postprocess(
295			self,
296			inputs: Dict[str, tf.Tensor],
297			outputs: Dict[str, tf.Tensor],
298			) -> Tuple[tf.Tensor, Dict]:
299			"""
300			Return a dict used for saving inputs and outputs.
301
302			:param inputs: dict of model inputs
303			:param outputs: dict of model outputs
304			:return: tuple, indices and a dict.
305			In the dict, each value is (tensor, normalize, on_label), where
306			- normalize = True if the tensor need to be normalized to [0, 1]
307			- on_label = True if the tensor depends on label
308			"""
309
310			def plot_model(self, output_dir: str):
311			"""
312			Save model structure in png.
313
314			:param output_dir: path to the output dir.
315			"""
316			logging.info(self._model.summary())
317			try:
318			tf.keras.utils.plot_model(
319			self._model,
320			to_file=os.path.join(output_dir, f"{self.name}.png"),
321			dpi=96,
322			show_shapes=True,
323			show_layer_names=True,
324			expand_nested=False,
325			)
326			except ImportError as err: # pragma: no cover
327			logging.error(
328			"Failed to plot model structure."
329			"Please check if graphviz is installed.\n"
330			"Error message is:"
331			f"{err}"
332			)
333
334			def log_tensor_stats(self, tensor: tf.Tensor, name: str):
335			"""
336			Log statistics of a given tensor.
337
338			:param tensor: tensor to monitor.
339			:param name: name of the tensor.
340			"""
341			flatten = tf.reshape(tensor, shape=(self.batch_size, -1))
342			self._model.add_metric(
343			tf.reduce_mean(flatten, axis=1),
344			name=f"metric/{name}_mean",
345			aggregation="mean",
346			)
347			self._model.add_metric(
348			tf.reduce_min(flatten, axis=1),
349			name=f"metric/{name}_min",
350			aggregation="min",
351			)
352			self._model.add_metric(
353			tf.reduce_max(flatten, axis=1),
354			name=f"metric/{name}_max",
355			aggregation="max",
356			)
357
358
359			@REGISTRY.register_model(name="ddf")
360			class DDFModel(RegistrationModel):
361			"""
362			A registration model predicts DDF.
363
364			When using global net as backbone,
365			the model predicts an affine transformation parameters,
366			and a DDF is calculated based on that.
367			"""
368
369			name = "DDFModel"
370
371			def _resize_interpolate(self, field, control_points):
372			resize = layer.ResizeCPTransform(control_points)
373			field = resize(field)
374
375			interpolate = layer.BSplines3DTransform(control_points, self.fixed_image_size)
376			field = interpolate(field)
377
378			return field
379
380			def build_model(self):
381			"""Build the model to be saved as self._model."""
382			# build inputs
383			self._inputs = self.build_inputs()
384			moving_image = self._inputs["moving_image"]
385			fixed_image = self._inputs["fixed_image"]
386
387			# build ddf
388			control_points = self.config["backbone"].pop("control_points", False)
389			backbone_inputs = self.concat_images(moving_image, fixed_image)
390			backbone = REGISTRY.build_backbone(
391			config=self.config["backbone"],
392			default_args=dict(
393			image_size=self.fixed_image_size,
394			out_channels=3,
395			out_kernel_initializer="zeros",
396			out_activation=None,
397			),
398			)
399
400			if isinstance(backbone, GlobalNet):
401			# (f_dim1, f_dim2, f_dim3, 3), (4, 3)
402			ddf, theta = backbone(inputs=backbone_inputs)
403			self._outputs = dict(ddf=ddf, theta=theta)
404			else:
405			# (f_dim1, f_dim2, f_dim3, 3)
406			ddf = backbone(inputs=backbone_inputs)
407			ddf = (
408			self._resize_interpolate(ddf, control_points) if control_points else ddf
409			)
410			self._outputs = dict(ddf=ddf)
411
412			# build outputs
413			warping = layer.Warping(fixed_image_size=self.fixed_image_size)
414			# (f_dim1, f_dim2, f_dim3, 3)
415			pred_fixed_image = warping(inputs=[ddf, moving_image])
416			self._outputs["pred_fixed_image"] = pred_fixed_image
417
418			if not self.labeled:
419			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
420
421			# (f_dim1, f_dim2, f_dim3, 3)
422			moving_label = self._inputs["moving_label"]
423			pred_fixed_label = warping(inputs=[ddf, moving_label])
424
425			self._outputs["pred_fixed_label"] = pred_fixed_label
426			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
427
428			def build_loss(self):
429			"""Build losses according to configs."""
430			super().build_loss()
431
432			# ddf loss and metrics
433			ddf = self._outputs["ddf"]
434			self._build_loss(name="regularization", inputs_dict=dict(inputs=ddf))
435			self.log_tensor_stats(tensor=ddf, name="ddf")
436
437			def postprocess(
438			self,
439			inputs: Dict[str, tf.Tensor],
440			outputs: Dict[str, tf.Tensor],
441			) -> Tuple[tf.Tensor, Dict]:
442			"""
443			Return a dict used for saving inputs and outputs.
444
445			:param inputs: dict of model inputs
446			:param outputs: dict of model outputs
447			:return: tuple, indices and a dict.
448			In the dict, each value is (tensor, normalize, on_label), where
449			- normalize = True if the tensor need to be normalized to [0, 1]
450			- on_label = True if the tensor depends on label
451			"""
452			indices = inputs["indices"]
453			processed = dict(
454			moving_image=(inputs["moving_image"], True, False),
455			fixed_image=(inputs["fixed_image"], True, False),
456			ddf=(outputs["ddf"], True, False),
457			pred_fixed_image=(outputs["pred_fixed_image"], True, False),
458			)
459
460			# save theta for affine model
461			if "theta" in outputs:
462			processed["theta"] = (outputs["theta"], None, None) # type: ignore
463
464			if not self.labeled:
465			return indices, processed
466
467			processed = {
468			**dict(
469			moving_label=(inputs["moving_label"], False, True),
470			fixed_label=(inputs["fixed_label"], False, True),
471			pred_fixed_label=(outputs["pred_fixed_label"], False, True),
472			),
473			**processed,
474			}
475
476			return indices, processed
477
478
479			@REGISTRY.register_model(name="dvf")
480			class DVFModel(DDFModel):
481			"""
482			A registration model predicts DVF.
483
484			DDF is calculated based on DVF.
485			"""
486
487			name = "DVFModel"
488
489			def build_model(self):
490			"""Build the model to be saved as self._model."""
491			# build inputs
492			self._inputs = self.build_inputs()
493			moving_image = self._inputs["moving_image"]
494			fixed_image = self._inputs["fixed_image"]
495			control_points = self.config["backbone"].pop("control_points", False)
496
497			# build ddf
498			backbone_inputs = self.concat_images(moving_image, fixed_image)
499			backbone = REGISTRY.build_backbone(
500			config=self.config["backbone"],
501			default_args=dict(
502			image_size=self.fixed_image_size,
503			out_channels=3,
504			out_kernel_initializer="zeros",
505			out_activation=None,
506			),
507			)
508			dvf = backbone(inputs=backbone_inputs)
509			dvf = self._resize_interpolate(dvf, control_points) if control_points else dvf
510			ddf = layer.IntDVF(fixed_image_size=self.fixed_image_size)(dvf)
511
512			# build outputs
513			self._warping = layer.Warping(fixed_image_size=self.fixed_image_size)
514			# (f_dim1, f_dim2, f_dim3, 3)
515			pred_fixed_image = self._warping(inputs=[ddf, moving_image])
516
517			self._outputs = dict(dvf=dvf, ddf=ddf, pred_fixed_image=pred_fixed_image)
518
519			if not self.labeled:
520			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
521
522			# (f_dim1, f_dim2, f_dim3, 3)
523			moving_label = self._inputs["moving_label"]
524			pred_fixed_label = self._warping(inputs=[ddf, moving_label])
525
526			self._outputs["pred_fixed_label"] = pred_fixed_label
527			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
528
529			def build_loss(self):
530			"""Build losses according to configs."""
531			super().build_loss()
532
533			# dvf metrics
534			dvf = self._outputs["dvf"]
535			self.log_tensor_stats(tensor=dvf, name="dvf")
536
537			def postprocess(
538			self,
539			inputs: Dict[str, tf.Tensor],
540			outputs: Dict[str, tf.Tensor],
541			) -> Tuple[tf.Tensor, Dict]:
542			"""
543			Return a dict used for saving inputs and outputs.
544
545			:param inputs: dict of model inputs
546			:param outputs: dict of model outputs
547			:return: tuple, indices and a dict.
548			In the dict, each value is (tensor, normalize, on_label), where
549			- normalize = True if the tensor need to be normalized to [0, 1]
550			- on_label = True if the tensor depends on label
551			"""
552			indices, processed = super().postprocess(inputs=inputs, outputs=outputs)
553			processed["dvf"] = (outputs["dvf"], True, False)
554			return indices, processed
555
556
557			@REGISTRY.register_model(name="conditional")
558			class ConditionalModel(RegistrationModel):
559			"""
560			A registration model predicts fixed image label without DDF or DVF.
561			"""
562
563			name = "ConditionalModel"
564
565			def build_model(self):
566			"""Build the model to be saved as self._model."""
567			assert self.labeled
568
569			# build inputs
570			self._inputs = self.build_inputs()
571			moving_image = self._inputs["moving_image"]
572			fixed_image = self._inputs["fixed_image"]
573			moving_label = self._inputs["moving_label"]
574
575			# build ddf
576			backbone_inputs = self.concat_images(moving_image, fixed_image, moving_label)
577			backbone = REGISTRY.build_backbone(
578			config=self.config["backbone"],
579			default_args=dict(
580			image_size=self.fixed_image_size,
581			out_channels=1,
582			out_kernel_initializer="glorot_uniform",
583			out_activation="sigmoid",
584			),
585			)
586			# (batch, f_dim1, f_dim2, f_dim3)
587			pred_fixed_label = backbone(inputs=backbone_inputs)
588			pred_fixed_label = tf.squeeze(pred_fixed_label, axis=4)
589
590			self._outputs = dict(pred_fixed_label=pred_fixed_label)
591			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
592
593			def postprocess(
594			self,
595			inputs: Dict[str, tf.Tensor],
596			outputs: Dict[str, tf.Tensor],
597			) -> Tuple[tf.Tensor, Dict]:
598			"""
599			Return a dict used for saving inputs and outputs.
600
601			:param inputs: dict of model inputs
602			:param outputs: dict of model outputs
603			:return: tuple, indices and a dict.
604			In the dict, each value is (tensor, normalize, on_label), where
605			- normalize = True if the tensor need to be normalized to [0, 1]
606			- on_label = True if the tensor depends on label
607			"""
608			indices = inputs["indices"]
609			processed = dict(
610			moving_image=(inputs["moving_image"], True, False),
611			fixed_image=(inputs["fixed_image"], True, False),
612			pred_fixed_label=(outputs["pred_fixed_label"], True, True),
613			moving_label=(inputs["moving_label"], False, True),
614			fixed_label=(inputs["fixed_label"], False, True),
615			)
616
617			return indices, processed
618

DeepRegNet / DeepReg

Push — main ( 3b5a47...3ab0c2 )

deepreg.model.network.DVFModel.build_loss() A

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like