deepreg.model.network.RegistrationModel._build_loss() - Code Metrics - Inspection of "Multiple loss functions" - DeepRegNet/DeepReg - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Pull Request — main (#662)

unknown

created 2021-02-01 14:02 UTC

RegistrationModel._build_loss() B

↳ Parent: deepreg.model.network

Complexity

Conditions

Size

Total Lines	56
Code Lines	29

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
eloc	29
dl	0
loc	56
rs	8.2506
c	0
b	0
f	0
cc	6
nop	3

How to fix Long Method

import logging

from abc import abstractmethod
from copy import deepcopy
from typing import Dict, Optional

import tensorflow as tf

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._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 = []

        # (batch, m_dim1, m_dim2, m_dim3, 1)
        moving_image = tf.expand_dims(moving_image, axis=4)
        moving_image = layer_util.resize3d(
            image=moving_image, size=self.fixed_image_size
        )
        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 = layer_util.resize3d(
                image=moving_label, size=self.fixed_image_size
            )
            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
        :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"Loss is not used."
            )
            return

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

        for loss_config in loss_list:

            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_cls = REGISTRY.build_loss(
                config=dict_without(d=loss_config, key="weight")
            )
            loss = loss_cls(**inputs_dict) / self.global_batch_size
            weighted_loss = loss * weight

            # add loss
            self._model.add_loss(weighted_loss)

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

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

    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],
    ) -> (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
        """


@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.
    """

    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
        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)
            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."""
        fixed_image = self._inputs["fixed_image"]
        ddf = self._outputs["ddf"]
        pred_fixed_image = self._outputs["pred_fixed_image"]

        # ddf
        self._build_loss(name="regularization", inputs_dict=dict(inputs=ddf))

        # image
        self._build_loss(
            name="image", inputs_dict=dict(y_true=fixed_image, y_pred=pred_fixed_image)
        )

        # label
        if self.labeled:
            fixed_label = self._inputs["fixed_label"]
            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),
            )

    def postprocess(
        self,
        inputs: Dict[str, tf.Tensor],
        outputs: Dict[str, tf.Tensor],
    ) -> (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)

        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.
    """

    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
        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)
        ddf = layer.IntDVF(fixed_image_size=self.fixed_image_size)(dvf)

        # 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 = 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 = warping(inputs=[ddf, moving_label])

        self._outputs["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],
    ) -> (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.
    """

    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 build_loss(self):
        """Build losses according to configs."""
        fixed_label = self._inputs["fixed_label"]
        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),
        )

    def postprocess(
        self,
        inputs: Dict[str, tf.Tensor],
        outputs: Dict[str, tf.Tensor],
    ) -> (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
			0 ignored issues – show introduced 2021-01-26 01:41 UTC by Report Bug Copy Issue Report Missing module docstring Loading history...
2			from abc import abstractmethod
3			from copy import deepcopy
4			from typing import Dict, Optional
5
6			import tensorflow as tf
7
8			from deepreg.model import layer, layer_util
9			from deepreg.model.backbone import GlobalNet
10			from deepreg.registry import REGISTRY
11
12
13			def dict_without(d: dict, key) -> dict:
			0 ignored issues – show introduced 2021-01-26 01:41 UTC by Report Bug Copy Issue Report "key" missing in parameter type documentation Loading history...
14			"""
15			Return a copy of the given dict without a certain key.
16
17			:param d: dict to be copied.
18			:param key: key to be removed.
19			:return: the copy without a key
20			"""
21			copied = deepcopy(d)
22			copied.pop(key)
23			return copied
24
25
26			class RegistrationModel(tf.keras.Model):
27			"""Interface for registration model."""
28
29			def __init__(
30			self,
31			moving_image_size: tuple,
32			fixed_image_size: tuple,
33			index_size: int,
34			labeled: bool,
35			batch_size: int,
36			config: dict,
37			num_devices: int = 1,
38			name: str = "RegistrationModel",
39			):
40			"""
41			Init.
42
43			:param moving_image_size: (m_dim1, m_dim2, m_dim3)
44			:param fixed_image_size: (f_dim1, f_dim2, f_dim3)
45			:param index_size: number of indices for identify each sample
46			:param labeled: if the data is labeled
47			:param batch_size: size of mini-batch
48			:param config: config for method, backbone, and loss.
49			:param num_devices: number of GPU used,
50			global_batch_size = batch_size*num_devices
51			:param name: name of the model
52			"""
53			super().__init__(name=name)
54			self.moving_image_size = moving_image_size
55			self.fixed_image_size = fixed_image_size
56			self.index_size = index_size
57			self.labeled = labeled
58			self.batch_size = batch_size
59			self.config = config
60			self.num_devices = num_devices
61			self.global_batch_size = num_devices * batch_size
62
63			self._inputs = None # save inputs of self._model as dict
64			self._outputs = None # save outputs of self._model as dict
65
66			self._model = self.build_model()
67			self.build_loss()
68
69			def get_config(self) -> dict:
70			"""Return the config dictionary for recreating this class."""
71			return dict(
72			moving_image_size=self.moving_image_size,
73			fixed_image_size=self.fixed_image_size,
74			index_size=self.index_size,
75			labeled=self.labeled,
76			batch_size=self.batch_size,
77			config=self.config,
78			num_devices=self.num_devices,
79			name=self.name,
80			)
81
82			@abstractmethod
83			def build_model(self):
84			"""Build the model to be saved as self._model."""
85
86			def build_inputs(self) -> Dict[str, tf.keras.layers.Input]:
87			"""
88			Build input tensors.
89
90			:return: dict of inputs.
91			"""
92			# (batch, m_dim1, m_dim2, m_dim3, 1)
93			moving_image = tf.keras.Input(
94			shape=self.moving_image_size,
95			batch_size=self.batch_size,
96			name="moving_image",
97			)
98			# (batch, f_dim1, f_dim2, f_dim3, 1)
99			fixed_image = tf.keras.Input(
100			shape=self.fixed_image_size,
101			batch_size=self.batch_size,
102			name="fixed_image",
103			)
104			# (batch, index_size)
105			indices = tf.keras.Input(
106			shape=(self.index_size,),
107			batch_size=self.batch_size,
108			name="indices",
109			)
110
111			if not self.labeled:
112			return dict(
113			moving_image=moving_image, fixed_image=fixed_image, indices=indices
114			)
115
116			# (batch, m_dim1, m_dim2, m_dim3, 1)
117			moving_label = tf.keras.Input(
118			shape=self.moving_image_size,
119			batch_size=self.batch_size,
120			name="moving_label",
121			)
122			# (batch, m_dim1, m_dim2, m_dim3, 1)
123			fixed_label = tf.keras.Input(
124			shape=self.fixed_image_size,
125			batch_size=self.batch_size,
126			name="fixed_label",
127			)
128			return dict(
129			moving_image=moving_image,
130			fixed_image=fixed_image,
131			moving_label=moving_label,
132			fixed_label=fixed_label,
133			indices=indices,
134			)
135
136			def concat_images(
137			self,
138			moving_image: tf.Tensor,
139			fixed_image: tf.Tensor,
140			moving_label: Optional[tf.Tensor] = None,
141			) -> tf.Tensor:
142			"""
143			Adjust image shape and concatenate them together.
144
145			:param moving_image: registration source
146			:param fixed_image: registration target
147			:param moving_label: optional, only used for conditional model.
148			:return:
149			"""
150			images = []
151
152			# (batch, m_dim1, m_dim2, m_dim3, 1)
153			moving_image = tf.expand_dims(moving_image, axis=4)
154			moving_image = layer_util.resize3d(
155			image=moving_image, size=self.fixed_image_size
156			)
157			images.append(moving_image)
158
159			# (batch, m_dim1, m_dim2, m_dim3, 1)
160			fixed_image = tf.expand_dims(fixed_image, axis=4)
161			images.append(fixed_image)
162
163			# (batch, m_dim1, m_dim2, m_dim3, 1)
164			if moving_label is not None:
165			moving_label = tf.expand_dims(moving_label, axis=4)
166			moving_label = layer_util.resize3d(
167			image=moving_label, size=self.fixed_image_size
168			)
169			images.append(moving_label)
170
171			# (batch, f_dim1, f_dim2, f_dim3, 2 or 3)
172			images = tf.concat(images, axis=4)
173			return images
174
175			def _build_loss(self, name: str, inputs_dict: dict):
176			"""
177			Build and add one weighted loss together with the metrics.
178
179			:param name: name of loss
180			:param inputs_dict: inputs for loss function
181			"""
182
183			if name not in self.config["loss"]:
184			# loss config is not defined
185			logging.warning(
			0 ignored issues – show introduced 2021-01-26 01:41 UTC by Report Bug Copy Issue Report Use lazy % formatting in logging functions Loading history...
186			f"The configuration for loss {name} is not defined."
187			f"Loss is not used."
188			)
189			return
190
191			loss_list = self.config["loss"][name]
192			if not isinstance(loss_list, list):
193			loss_list = [loss_list]
194
195			for loss_config in loss_list:
196
197			if "weight" not in loss_config:
198			# default loss weight 1
199			logging.warning(
			0 ignored issues – show introduced 2021-01-26 01:41 UTC by Report Bug Copy Issue Report Use lazy % formatting in logging functions Loading history...
200			f"The weight for loss {name} is not defined."
201			f"Default weight = 1.0 is used."
202			)
203			loss_config["weight"] = 1.0
204
205			# build loss
206			weight = loss_config["weight"]
207
208			if weight == 0:
209			logging.warning(
			0 ignored issues – show introduced 2021-02-01 14:03 UTC by Report Bug Copy Issue Report Use lazy % formatting in logging functions Loading history...
210			f"The weight for loss {name} is zero." f"Loss is not used."
211			)
212			return
213
214			loss_cls = REGISTRY.build_loss(
215			config=dict_without(d=loss_config, key="weight")
216			)
217			loss = loss_cls(**inputs_dict) / self.global_batch_size
218			weighted_loss = loss * weight
219
220			# add loss
221			self._model.add_loss(weighted_loss)
222
223			# add metric
224			self._model.add_metric(
225			loss, name=f"loss/{name}_{loss_cls.name}", aggregation="mean"
226			)
227			self._model.add_metric(
228			weighted_loss,
229			name=f"loss/{name}_{loss_cls.name}_weighted",
230			aggregation="mean",
231			)
232
233			@abstractmethod
234			def build_loss(self):
235			"""Build losses according to configs."""
236
237			def call(
			0 ignored issues – show introduced 2021-01-26 01:41 UTC by Report Bug Copy Issue Report "mask, training" missing in parameter type documentation Loading history...
238			self, inputs: Dict[str, tf.Tensor], training=None, mask=None
239			) -> Dict[str, tf.Tensor]:
240			"""
241			Call the self._model.
242
243			:param inputs: a dict of tensors.
244			:param training: training or not.
245			:param mask: maks for inputs.
246			:return:
247			"""
248			return self._model(inputs, training=training, mask=mask) # pragma: no cover
249
250			@abstractmethod
251			def postprocess(
252			self,
253			inputs: Dict[str, tf.Tensor],
254			outputs: Dict[str, tf.Tensor],
255			) -> (tf.Tensor, Dict):
256			"""
257			Return a dict used for saving inputs and outputs.
258
259			:param inputs: dict of model inputs
260			:param outputs: dict of model outputs
261			:return: tuple, indices and a dict.
262			In the dict, each value is (tensor, normalize, on_label), where
263			- normalize = True if the tensor need to be normalized to [0, 1]
264			- on_label = True if the tensor depends on label
265			"""
266
267
268			@REGISTRY.register_model(name="ddf")
269			class DDFModel(RegistrationModel):
270			"""
271			A registration model predicts DDF.
272
273			When using global net as backbone,
274			the model predicts an affine transformation parameters,
275			and a DDF is calculated based on that.
276			"""
277
278			def build_model(self):
279			"""Build the model to be saved as self._model."""
280			# build inputs
281			self._inputs = self.build_inputs()
282			moving_image = self._inputs["moving_image"]
283			fixed_image = self._inputs["fixed_image"]
284
285			# build ddf
286			backbone_inputs = self.concat_images(moving_image, fixed_image)
287			backbone = REGISTRY.build_backbone(
288			config=self.config["backbone"],
289			default_args=dict(
290			image_size=self.fixed_image_size,
291			out_channels=3,
292			out_kernel_initializer="zeros",
293			out_activation=None,
294			),
295			)
296
297			if isinstance(backbone, GlobalNet):
298			# (f_dim1, f_dim2, f_dim3, 3), (4, 3)
299			ddf, theta = backbone(inputs=backbone_inputs)
300			self._outputs = dict(ddf=ddf, theta=theta)
301			else:
302			# (f_dim1, f_dim2, f_dim3, 3)
303			ddf = backbone(inputs=backbone_inputs)
304			self._outputs = dict(ddf=ddf)
305
306			# build outputs
307			warping = layer.Warping(fixed_image_size=self.fixed_image_size)
308			# (f_dim1, f_dim2, f_dim3, 3)
309			pred_fixed_image = warping(inputs=[ddf, moving_image])
310			self._outputs["pred_fixed_image"] = pred_fixed_image
311
312			if not self.labeled:
313			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
314
315			# (f_dim1, f_dim2, f_dim3, 3)
316			moving_label = self._inputs["moving_label"]
317			pred_fixed_label = warping(inputs=[ddf, moving_label])
318
319			self._outputs["pred_fixed_label"] = pred_fixed_label
320			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
321
322			def build_loss(self):
323			"""Build losses according to configs."""
324			fixed_image = self._inputs["fixed_image"]
325			ddf = self._outputs["ddf"]
326			pred_fixed_image = self._outputs["pred_fixed_image"]
327
328			# ddf
329			self._build_loss(name="regularization", inputs_dict=dict(inputs=ddf))
330
331			# image
332			self._build_loss(
333			name="image", inputs_dict=dict(y_true=fixed_image, y_pred=pred_fixed_image)
334			)
335
336			# label
337			if self.labeled:
338			fixed_label = self._inputs["fixed_label"]
339			pred_fixed_label = self._outputs["pred_fixed_label"]
340			self._build_loss(
341			name="label",
342			inputs_dict=dict(y_true=fixed_label, y_pred=pred_fixed_label),
343			)
344
345			def postprocess(
346			self,
347			inputs: Dict[str, tf.Tensor],
348			outputs: Dict[str, tf.Tensor],
349			) -> (tf.Tensor, Dict):
350			"""
351			Return a dict used for saving inputs and outputs.
352
353			:param inputs: dict of model inputs
354			:param outputs: dict of model outputs
355			:return: tuple, indices and a dict.
356			In the dict, each value is (tensor, normalize, on_label), where
357			- normalize = True if the tensor need to be normalized to [0, 1]
358			- on_label = True if the tensor depends on label
359			"""
360			indices = inputs["indices"]
361			processed = dict(
362			moving_image=(inputs["moving_image"], True, False),
363			fixed_image=(inputs["fixed_image"], True, False),
364			ddf=(outputs["ddf"], True, False),
365			pred_fixed_image=(outputs["pred_fixed_image"], True, False),
366			)
367
368			# save theta for affine model
369			if "theta" in outputs:
370			processed["theta"] = (outputs["theta"], None, None)
371
372			if not self.labeled:
373			return indices, processed
374
375			processed = {
376			**dict(
377			moving_label=(inputs["moving_label"], False, True),
378			fixed_label=(inputs["fixed_label"], False, True),
379			pred_fixed_label=(outputs["pred_fixed_label"], False, True),
380			),
381			**processed,
382			}
383
384			return indices, processed
385
386
387			@REGISTRY.register_model(name="dvf")
388			class DVFModel(DDFModel):
389			"""
390			A registration model predicts DVF.
391
392			DDF is calculated based on DVF.
393			"""
394
395			def build_model(self):
396			"""Build the model to be saved as self._model."""
397			# build inputs
398			self._inputs = self.build_inputs()
399			moving_image = self._inputs["moving_image"]
400			fixed_image = self._inputs["fixed_image"]
401
402			# build ddf
403			backbone_inputs = self.concat_images(moving_image, fixed_image)
404			backbone = REGISTRY.build_backbone(
405			config=self.config["backbone"],
406			default_args=dict(
407			image_size=self.fixed_image_size,
408			out_channels=3,
409			out_kernel_initializer="zeros",
410			out_activation=None,
411			),
412			)
413			dvf = backbone(inputs=backbone_inputs)
414			ddf = layer.IntDVF(fixed_image_size=self.fixed_image_size)(dvf)
415
416			# build outputs
417			warping = layer.Warping(fixed_image_size=self.fixed_image_size)
418			# (f_dim1, f_dim2, f_dim3, 3)
419			pred_fixed_image = warping(inputs=[ddf, moving_image])
420
421			self._outputs = dict(dvf=dvf, ddf=ddf, pred_fixed_image=pred_fixed_image)
422
423			if not self.labeled:
424			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
425
426			# (f_dim1, f_dim2, f_dim3, 3)
427			moving_label = self._inputs["moving_label"]
428			pred_fixed_label = warping(inputs=[ddf, moving_label])
429
430			self._outputs["pred_fixed_label"] = pred_fixed_label
431			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
432
433			def postprocess(
434			self,
435			inputs: Dict[str, tf.Tensor],
436			outputs: Dict[str, tf.Tensor],
437			) -> (tf.Tensor, Dict):
438			"""
439			Return a dict used for saving inputs and outputs.
440
441			:param inputs: dict of model inputs
442			:param outputs: dict of model outputs
443			:return: tuple, indices and a dict.
444			In the dict, each value is (tensor, normalize, on_label), where
445			- normalize = True if the tensor need to be normalized to [0, 1]
446			- on_label = True if the tensor depends on label
447			"""
448			indices, processed = super().postprocess(inputs=inputs, outputs=outputs)
449			processed["dvf"] = (outputs["dvf"], True, False)
450			return indices, processed
451
452
453			@REGISTRY.register_model(name="conditional")
454			class ConditionalModel(RegistrationModel):
455			"""
456			A registration model predicts fixed image label without DDF or DVF.
457			"""
458
459			def build_model(self):
460			"""Build the model to be saved as self._model."""
461			assert self.labeled
462
463			# build inputs
464			self._inputs = self.build_inputs()
465			moving_image = self._inputs["moving_image"]
466			fixed_image = self._inputs["fixed_image"]
467			moving_label = self._inputs["moving_label"]
468
469			# build ddf
470			backbone_inputs = self.concat_images(moving_image, fixed_image, moving_label)
471			backbone = REGISTRY.build_backbone(
472			config=self.config["backbone"],
473			default_args=dict(
474			image_size=self.fixed_image_size,
475			out_channels=1,
476			out_kernel_initializer="glorot_uniform",
477			out_activation="sigmoid",
478			),
479			)
480			# (batch, f_dim1, f_dim2, f_dim3)
481			pred_fixed_label = backbone(inputs=backbone_inputs)
482			pred_fixed_label = tf.squeeze(pred_fixed_label, axis=4)
483
484			self._outputs = dict(pred_fixed_label=pred_fixed_label)
485			return tf.keras.Model(inputs=self._inputs, outputs=self._outputs)
486
487			def build_loss(self):
488			"""Build losses according to configs."""
489			fixed_label = self._inputs["fixed_label"]
490			pred_fixed_label = self._outputs["pred_fixed_label"]
491
492			self._build_loss(
493			name="label",
494			inputs_dict=dict(y_true=fixed_label, y_pred=pred_fixed_label),
495			)
496
497			def postprocess(
498			self,
499			inputs: Dict[str, tf.Tensor],
500			outputs: Dict[str, tf.Tensor],
501			) -> (tf.Tensor, Dict):
502			"""
503			Return a dict used for saving inputs and outputs.
504
505			:param inputs: dict of model inputs
506			:param outputs: dict of model outputs
507			:return: tuple, indices and a dict.
508			In the dict, each value is (tensor, normalize, on_label), where
509			- normalize = True if the tensor need to be normalized to [0, 1]
510			- on_label = True if the tensor depends on label
511			"""
512			indices = inputs["indices"]
513			processed = dict(
514			moving_image=(inputs["moving_image"], True, False),
515			fixed_image=(inputs["fixed_image"], True, False),
516			pred_fixed_label=(outputs["pred_fixed_label"], True, True),
517			moving_label=(inputs["moving_label"], False, True),
518			fixed_label=(inputs["fixed_label"], False, True),
519			)
520
521			return indices, processed
522

DeepRegNet / DeepReg

Pull Request — main (#662)

RegistrationModel._build_loss() B

Complexity

Size

Duplication

Importance

How to fix Long Method

Long Method

Duplication Side-by-Side

Filter issues like