deepreg.model.backbone.local_net.AdditiveUpsampling.__init__() - Code Metrics - Inspection of "640 refactor model layers" - DeepRegNet/DeepReg - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Pull Request — main (#656)

by Yunguan

created 2021-03-04 22:39 UTC

AdditiveUpsampling.init() A

↳ Parent: deepreg.model.backbone.local_net

Complexity

Conditions

Size

Total Lines	30
Code Lines	17

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
eloc	17
dl	0
loc	30
rs	9.55
c	0
b	0
f	0
cc	1
nop	8

How to fix Many Parameters

# coding=utf-8


from typing import List, Optional, Tuple, Union

import tensorflow as tf
import tensorflow.keras.layers as tfkl

from deepreg.model import layer
from deepreg.model.backbone.u_net import UNet
from deepreg.model.layer import Extraction
from deepreg.registry import REGISTRY


class AdditiveUpsampling(tfkl.Layer):

    def __init__(
        self,
        filters: int,
        output_padding: Union[int, Tuple, List],
        kernel_size: Union[int, Tuple, List],
        padding: str,
        strides: Union[int, Tuple, List],
        output_shape: Tuple,
        name: str = "AdditiveUpsampling",
    ):
        """
        Addictive up-sampling layer.

        :param filters: number of channels for output
        :param output_padding: padding for output
        :param kernel_size: arg for deconv3d
        :param padding: arg for deconv3d
        :param strides: arg for deconv3d
        :param output_shape: shape of the output tensor
        :param name: name of the layer.
        """
        super().__init__(name=name)
        self.deconv3d = layer.Deconv3dBlock(
            filters=filters,
            output_padding=output_padding,
            kernel_size=kernel_size,
            strides=strides,
            padding=padding,
        )
        self.resize = layer.Resize3d(shape=output_shape)

    def call(self, inputs, **kwargs):
        deconved = self.deconv3d(inputs)
        resized = self.resize(inputs)
        resized = tf.add_n(tf.split(resized, num_or_size_splits=2, axis=4))

        return deconved + resized

    def get_config(self) -> dict:
        """Return the config dictionary for recreating this class."""
        config = super().get_config()
        deconv_config = self.deconv3d.get_config()
        config.update(
            filters=deconv_config["filters"],
            output_padding=deconv_config["output_padding"],
            kernel_size=deconv_config["kernel_size"],
            strides=deconv_config["strides"],
            padding=deconv_config["padding"],
        )
        config.update(output_shape=self.resize._shape)
class MyParent:
    def __init__(self):
        self._x = 1;
        self.y = 2;

class MyChild(MyParent):
    def some_method(self):
        return self._x    # Ok, since accessed from a child class

class AnotherClass:
    def some_method(self, instance_of_my_child):
        return instance_of_my_child._x   # Would be flagged as AnotherClass is not
                                         # a child class of MyParent
        return config


@REGISTRY.register_backbone(name="local")
class LocalNet(UNet):
    """
    Build LocalNet for image registration.

    Reference:

    - Hu, Yipeng, et al.
      "Weakly-supervised convolutional neural networks
      for multimodal image registration."
      Medical image analysis 49 (2018): 1-13.
      https://doi.org/10.1016/j.media.2018.07.002

    - Hu, Yipeng, et al.
      "Label-driven weakly-supervised learning
      for multimodal deformable image registration,"
      https://arxiv.org/abs/1711.01666
    """

    def __init__(
        self,
        image_size: tuple,
        num_channel_initial: int,
        extract_levels: Tuple[int],
        out_kernel_initializer: str,
        out_activation: str,
        out_channels: int,
        depth: Optional[int] = None,
        use_additive_upsampling: bool = True,
        pooling: bool = True,
        concat_skip: bool = False,
        name: str = "LocalNet",
        **kwargs,
    ):
        """
        Init.

        Image is encoded gradually, i from level 0 to D,
        then it is decoded gradually, j from level D to 0.
        Some of the decoded levels are used for generating extractions.

        So, extract_levels are between [0, D].

        :param image_size: such as (dim1, dim2, dim3)
        :param num_channel_initial: number of initial channels.
        :param extract_levels: from which depths the output will be built.
        :param out_kernel_initializer: initializer to use for kernels.
        :param out_activation: activation to use at end layer.
        :param out_channels: number of channels for the extractions
        :param depth: depth of the encoder.
            If depth is not given, depth = max(extract_levels) will be used.
        :param use_additive_upsampling: whether use additive up-sampling layer
            for decoding.
        :param pooling: for down-sampling, use non-parameterized
                        pooling if true, otherwise use conv3d
        :param concat_skip: when up-sampling, concatenate skipped
                            tensor if true, otherwise use addition
        :param name: name of the backbone.
        :param kwargs: additional arguments.
        """
        self._use_additive_upsampling = use_additive_upsampling
        if depth is None:
            depth = max(extract_levels)
        kwargs["encode_kernel_sizes"] = [7] + [3] * depth
        super().__init__(
            image_size=image_size,
            num_channel_initial=num_channel_initial,
            depth=depth,
            extract_levels=extract_levels,
            out_kernel_initializer=out_kernel_initializer,
            out_activation=out_activation,
            out_channels=out_channels,
            pooling=pooling,
            concat_skip=concat_skip,
            name=name,
            **kwargs,
        )

    def build_bottom_block(
        self, filters: int, kernel_size: int, padding: str
    ) -> Union[tf.keras.Model, tfkl.Layer]:
        """
        Build a block for bottom layer.

        This block do not change the tensor shape (width, height, depth),
        it only changes the number of channels.

        :param filters: number of channels for output
        :param kernel_size: arg for conv3d
        :param padding: arg for conv3d
        :return: a block consists of one or multiple layers
        """
        return layer.Conv3dBlock(
            filters=filters, kernel_size=kernel_size, padding=padding
        )

    def build_up_sampling_block(
        self,
        filters: int,
        output_padding: int,
        kernel_size: int,
        padding: str,
        strides: int,
        output_shape: tuple,
    ) -> Union[tf.keras.Model, tfkl.Layer]:
        """
        Build a block for up-sampling.

        This block changes the tensor shape (width, height, depth),
        but it does not changes the number of channels.

        :param filters: number of channels for output
        :param output_padding: padding for output
        :param kernel_size: arg for deconv3d
        :param padding: arg for deconv3d
        :param strides: arg for deconv3d
        :param output_shape: shape of the output tensor
        :return: a block consists of one or multiple layers
        """

        if self._use_additive_upsampling:
            return AdditiveUpsampling(
                filters=filters,
                output_padding=output_padding,
                kernel_size=kernel_size,
                strides=strides,
                padding=padding,
                output_shape=output_shape,
            )

        return layer.Deconv3dBlock(
            filters=filters,
            output_padding=output_padding,
            kernel_size=kernel_size,
            strides=strides,
            padding=padding,
        )

    def build_output_block(
        self,
        image_size: Tuple[int],
        extract_levels: Tuple[int],
        out_channels: int,
        out_kernel_initializer: str,
        out_activation: str,
    ) -> Union[tf.keras.Model, tfkl.Layer]:
        """
        Build a block for output.

        The input to this block is a list of tensors.

        :param image_size: such as (dim1, dim2, dim3)
        :param extract_levels: number of extraction levels.
        :param out_channels: number of channels for the extractions
        :param out_kernel_initializer: initializer to use for kernels.
        :param out_activation: activation to use at end layer.
        :return: a block consists of one or multiple layers
        """
        return Extraction(
            image_size=image_size,
            extract_levels=extract_levels,
            out_channels=out_channels,
            out_kernel_initializer=out_kernel_initializer,
            out_activation=out_activation,
        )

    def get_config(self) -> dict:
        """Return the config dictionary for recreating this class."""
        config = super().get_config()
        config.update(use_additive_upsampling=self._use_additive_upsampling)
        return config


1			# coding=utf-8
			0 ignored issues – show introduced 2021-01-26 01:41 UTC by Report Bug Copy Issue Report Missing module docstring Loading history...
2
3			from typing import List, Optional, Tuple, Union
4
5			import tensorflow as tf
6			import tensorflow.keras.layers as tfkl
7
8			from deepreg.model import layer
9			from deepreg.model.backbone.u_net import UNet
10			from deepreg.model.layer import Extraction
11			from deepreg.registry import REGISTRY
12
13
14			class AdditiveUpsampling(tfkl.Layer):
			0 ignored issues – show introduced 2021-01-29 00:08 UTC by Report Bug Copy Issue Report Missing class docstring Loading history...
15			def __init__(
16			self,
17			filters: int,
18			output_padding: Union[int, Tuple, List],
19			kernel_size: Union[int, Tuple, List],
20			padding: str,
21			strides: Union[int, Tuple, List],
22			output_shape: Tuple,
23			name: str = "AdditiveUpsampling",
24			):
25			"""
26			Addictive up-sampling layer.
27
28			:param filters: number of channels for output
29			:param output_padding: padding for output
30			:param kernel_size: arg for deconv3d
31			:param padding: arg for deconv3d
32			:param strides: arg for deconv3d
33			:param output_shape: shape of the output tensor
34			:param name: name of the layer.
35			"""
36			super().__init__(name=name)
37			self.deconv3d = layer.Deconv3dBlock(
38			filters=filters,
39			output_padding=output_padding,
40			kernel_size=kernel_size,
41			strides=strides,
42			padding=padding,
43			)
44			self.resize = layer.Resize3d(shape=output_shape)
45
46			def call(self, inputs, **kwargs):
47			deconved = self.deconv3d(inputs)
48			resized = self.resize(inputs)
49			resized = tf.add_n(tf.split(resized, num_or_size_splits=2, axis=4))
			0 ignored issues – show Unused Code introduced 2021-02-17 00:58 UTC by Report Bug Copy Issue Report Argument 'axis' passed by position and keyword in function call Loading history...
50			return deconved + resized
51
52			def get_config(self) -> dict:
53			"""Return the config dictionary for recreating this class."""
54			config = super().get_config()
55			deconv_config = self.deconv3d.get_config()
56			config.update(
57			filters=deconv_config["filters"],
58			output_padding=deconv_config["output_padding"],
59			kernel_size=deconv_config["kernel_size"],
60			strides=deconv_config["strides"],
61			padding=deconv_config["padding"],
62			)
63			config.update(output_shape=self.resize._shape)
			0 ignored issues – show Coding Style Best Practice introduced 2021-02-20 14:46 UTC by Report Bug Copy Issue Report It seems like `_shape` was declared protected and should not be accessed from this context. Prefixing a member variable `_` is usually regarded as the equivalent of declaring it with protected visibility that exists in other languages. Consequentially, such a member should only be accessed from the same class or a child class: class MyParent: def __init__(self): self._x = 1; self.y = 2; class MyChild(MyParent): def some_method(self): return self._x # Ok, since accessed from a child class class AnotherClass: def some_method(self, instance_of_my_child): return instance_of_my_child._x # Would be flagged as AnotherClass is not # a child class of MyParent Loading history...
64			return config
65
66
67			@REGISTRY.register_backbone(name="local")
68			class LocalNet(UNet):
69			"""
70			Build LocalNet for image registration.
71
72			Reference:
73
74			- Hu, Yipeng, et al.
75			"Weakly-supervised convolutional neural networks
76			for multimodal image registration."
77			Medical image analysis 49 (2018): 1-13.
78			https://doi.org/10.1016/j.media.2018.07.002
79
80			- Hu, Yipeng, et al.
81			"Label-driven weakly-supervised learning
82			for multimodal deformable image registration,"
83			https://arxiv.org/abs/1711.01666
84			"""
85
86			def __init__(
87			self,
88			image_size: tuple,
89			num_channel_initial: int,
90			extract_levels: Tuple[int],
91			out_kernel_initializer: str,
92			out_activation: str,
93			out_channels: int,
94			depth: Optional[int] = None,
95			use_additive_upsampling: bool = True,
96			pooling: bool = True,
97			concat_skip: bool = False,
98			name: str = "LocalNet",
99			**kwargs,
100			):
101			"""
102			Init.
103
104			Image is encoded gradually, i from level 0 to D,
105			then it is decoded gradually, j from level D to 0.
106			Some of the decoded levels are used for generating extractions.
107
108			So, extract_levels are between [0, D].
109
110			:param image_size: such as (dim1, dim2, dim3)
111			:param num_channel_initial: number of initial channels.
112			:param extract_levels: from which depths the output will be built.
113			:param out_kernel_initializer: initializer to use for kernels.
114			:param out_activation: activation to use at end layer.
115			:param out_channels: number of channels for the extractions
116			:param depth: depth of the encoder.
117			If depth is not given, depth = max(extract_levels) will be used.
118			:param use_additive_upsampling: whether use additive up-sampling layer
119			for decoding.
120			:param pooling: for down-sampling, use non-parameterized
121			pooling if true, otherwise use conv3d
122			:param concat_skip: when up-sampling, concatenate skipped
123			tensor if true, otherwise use addition
124			:param name: name of the backbone.
125			:param kwargs: additional arguments.
126			"""
127			self._use_additive_upsampling = use_additive_upsampling
128			if depth is None:
129			depth = max(extract_levels)
130			kwargs["encode_kernel_sizes"] = [7] + [3] * depth
131			super().__init__(
132			image_size=image_size,
133			num_channel_initial=num_channel_initial,
134			depth=depth,
135			extract_levels=extract_levels,
136			out_kernel_initializer=out_kernel_initializer,
137			out_activation=out_activation,
138			out_channels=out_channels,
139			pooling=pooling,
140			concat_skip=concat_skip,
141			name=name,
142			**kwargs,
143			)
144
145			def build_bottom_block(
146			self, filters: int, kernel_size: int, padding: str
147			) -> Union[tf.keras.Model, tfkl.Layer]:
148			"""
149			Build a block for bottom layer.
150
151			This block do not change the tensor shape (width, height, depth),
152			it only changes the number of channels.
153
154			:param filters: number of channels for output
155			:param kernel_size: arg for conv3d
156			:param padding: arg for conv3d
157			:return: a block consists of one or multiple layers
158			"""
159			return layer.Conv3dBlock(
160			filters=filters, kernel_size=kernel_size, padding=padding
161			)
162
163			def build_up_sampling_block(
164			self,
165			filters: int,
166			output_padding: int,
167			kernel_size: int,
168			padding: str,
169			strides: int,
170			output_shape: tuple,
171			) -> Union[tf.keras.Model, tfkl.Layer]:
172			"""
173			Build a block for up-sampling.
174
175			This block changes the tensor shape (width, height, depth),
176			but it does not changes the number of channels.
177
178			:param filters: number of channels for output
179			:param output_padding: padding for output
180			:param kernel_size: arg for deconv3d
181			:param padding: arg for deconv3d
182			:param strides: arg for deconv3d
183			:param output_shape: shape of the output tensor
184			:return: a block consists of one or multiple layers
185			"""
186
187			if self._use_additive_upsampling:
188			return AdditiveUpsampling(
189			filters=filters,
190			output_padding=output_padding,
191			kernel_size=kernel_size,
192			strides=strides,
193			padding=padding,
194			output_shape=output_shape,
195			)
196
197			return layer.Deconv3dBlock(
198			filters=filters,
199			output_padding=output_padding,
200			kernel_size=kernel_size,
201			strides=strides,
202			padding=padding,
203			)
204
205			def build_output_block(
206			self,
207			image_size: Tuple[int],
208			extract_levels: Tuple[int],
209			out_channels: int,
210			out_kernel_initializer: str,
211			out_activation: str,
212			) -> Union[tf.keras.Model, tfkl.Layer]:
213			"""
214			Build a block for output.
215
216			The input to this block is a list of tensors.
217
218			:param image_size: such as (dim1, dim2, dim3)
219			:param extract_levels: number of extraction levels.
220			:param out_channels: number of channels for the extractions
221			:param out_kernel_initializer: initializer to use for kernels.
222			:param out_activation: activation to use at end layer.
223			:return: a block consists of one or multiple layers
224			"""
225			return Extraction(
226			image_size=image_size,
227			extract_levels=extract_levels,
228			out_channels=out_channels,
229			out_kernel_initializer=out_kernel_initializer,
230			out_activation=out_activation,
231			)
232
233			def get_config(self) -> dict:
234			"""Return the config dictionary for recreating this class."""
235			config = super().get_config()
236			config.update(use_additive_upsampling=self._use_additive_upsampling)
237			return config
238

DeepRegNet / DeepReg

Pull Request — main (#656)

AdditiveUpsampling.__init__() A

Complexity

Size

Duplication

Importance

How to fix Many Parameters

Many Parameters

Duplication Side-by-Side

Filter issues like

AdditiveUpsampling.init() A