efficient_net.EfficientNet.build_efficient_net() - Code Metrics - Inspection of "Issue 678: Add EfficientNet" - DeepRegNet/DeepReg - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Pull Request — main (#785)

unknown

created 2021-06-21 15:06 UTC

efficient_net.EfficientNet.build_efficient_net() A

↳ Parent: efficient_net

Complexity

Conditions

Size

Total Lines	51
Code Lines	35

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
eloc	35
dl	0
loc	51
rs	9.0399
c	0
b	0
f	0
cc	4
nop	3

How to fix Long Method

"""This script provides an example of using efficient for training."""

import os
import math
import numpy as np
import tensorflow as tf
from tensorflow.keras import layers
from copy import deepcopy
from typing import List, Optional, Tuple, Union

from deepreg.model import layer
from deepreg.model.backbone import Backbone
from deepreg.model.backbone.local_net import LocalNet
from deepreg.model.backbone.u_net import UNet
from deepreg.model.layer import Extraction
from deepreg.registry import REGISTRY
from deepreg.train import train


EFFICIENTNET_PARAMS = {
    # model_name: (width_mult, depth_mult, image_size, dropout_rate, dropconnect_rate)
    "efficientnet-b0": (1.0, 1.0, 224, 0.2, 0.2),
    "efficientnet-b1": (1.0, 1.1, 240, 0.2, 0.2),
    "efficientnet-b2": (1.1, 1.2, 260, 0.3, 0.2),
    "efficientnet-b3": (1.2, 1.4, 300, 0.3, 0.2),
    "efficientnet-b4": (1.4, 1.8, 380, 0.4, 0.2),
    "efficientnet-b5": (1.6, 2.2, 456, 0.4, 0.2),
    "efficientnet-b6": (1.8, 2.6, 528, 0.5, 0.2),
    "efficientnet-b7": (2.0, 3.1, 600, 0.5, 0.2),
}

# Each Blocks Parameters
DEFAULT_BLOCKS_ARGS = [
    {'kernel_size': 3, 'repeats': 1, 'filters_in': 32, 'filters_out': 16,
     'expand_ratio': 1, 'id_skip': True, 'strides': 1, 'se_ratio': 0.25},
    {'kernel_size': 3, 'repeats': 2, 'filters_in': 16, 'filters_out': 24,
     'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
    {'kernel_size': 5, 'repeats': 2, 'filters_in': 24, 'filters_out': 40,
     'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
    {'kernel_size': 3, 'repeats': 3, 'filters_in': 40, 'filters_out': 80,
     'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
    {'kernel_size': 5, 'repeats': 3, 'filters_in': 80, 'filters_out': 112,
     'expand_ratio': 6, 'id_skip': True, 'strides': 1, 'se_ratio': 0.25},
    {'kernel_size': 5, 'repeats': 4, 'filters_in': 112, 'filters_out': 192,
     'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
    {'kernel_size': 3, 'repeats': 1, 'filters_in': 192, 'filters_out': 320,
     'expand_ratio': 6, 'id_skip': True, 'strides': 1, 'se_ratio': 0.25}
]

# Two Kernel Initializer
CONV_KERNEL_INITIALIZER = {
    'class_name': 'VarianceScaling',
    'config': {
        'scale': 2.0,
        'mode': 'fan_out',
        'distribution': 'normal'
    }
}

DENSE_KERNEL_INITIALIZER = {
    'class_name': 'VarianceScaling',
    'config': {
        'scale': 1. / 3.,
        'mode': 'fan_out',
        'distribution': 'uniform'
    }
}


@REGISTRY.register_backbone(name="efficient_net")
class EfficientNet(LocalNet):
    """
    Class that implements an Efficient-Net for image registration.

    Reference:
    - Author: Mingxing Tan, Quoc V. Le,
      EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
      https://arxiv.org/pdf/1905.11946.pdf
    """
    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,
        pooling: bool = True,
        concat_skip: bool = False,
        width_coefficient: float = 1.0,
        depth_coefficient: float = 1.0,
        default_size: int = 224,
        dropout_rate: float = 0.2,
        drop_connect_rate: float = 0.2,
        depth_divisor: int = 8,
        name: str = "EfficientNet",
        **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 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 width_coefficient: float, scaling coefficient for network width.
        :param depth_coefficient: float, scaling coefficient for network depth.
        :param default_size: int, default input image size.
        :param dropout_rate: float, dropout rate before final classifier layer.
        :param drop_connect_rate: float, dropout rate at skip connections.
        :param depth_divisor: int divisor for depth.
        :param name: name of the backbone.
        :param kwargs: additional arguments.
        """
        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,
            use_additive_upsampling = False,
            pooling=pooling,
            concat_skip=concat_skip,
            name=name,
            **kwargs,
        )

        self.width_coefficient =  width_coefficient
        self.depth_coefficient = depth_coefficient
        self.default_size = default_size
        self.dropout_rate = dropout_rate
        self.drop_connect_rate = drop_connect_rate
        self.depth_divisor = depth_divisor
        self.activation_fn = tf.nn.swish

    def call(self, inputs: tf.Tensor, training=None, mask=None) -> tf.Tensor:

        """
        Build compute graph based on built layers.

        :param inputs: image batch, shape = (batch, f_dim1, f_dim2, f_dim3, ch)
        :param training: None or bool.
        :param mask: None or tf.Tensor.
        :return: shape = (batch, f_dim1, f_dim2, f_dim3, out_channels)
        """

        # encoding / down-sampling
        skips = []
        encoded = inputs
        for d in range(self._depth):
            skip = self._encode_convs[d](inputs=encoded, training=training)
            encoded = self._encode_pools[d](inputs=skip, training=training)
            skips.append(skip)

        # bottom
        decoded = self.build_efficient_net(inputs=encoded, training=training)  # type: ignore

        # decoding / up-sampling
        outs = [decoded]
        for d in range(self._depth - 1, min(self._extract_levels) - 1, -1):
            decoded = self._decode_deconvs[d](inputs=decoded, training=training)
            decoded = self.build_skip_block()([decoded, skips[d]])
            decoded = self._decode_convs[d](inputs=decoded, training=training)
            outs = [decoded] + outs

        # output
        output = self._output_block(outs)  # type: ignore

        return output


    def build_efficient_net(self, inputs: tf.Tensor, training=None) -> tf.Tensor:
        """
        Builds graph based on built layers.

        :param inputs: shape = (batch, f_dim1, f_dim2, f_dim3, in_channels)
        :param training:
        :param mask:
        :return: shape = (batch, f_dim1, f_dim2, f_dim3, out_channels)
        """
        img_input = layers.Input(tensor=inputs, shape=self.image_size)
        bn_axis = 4  
        x = img_input
        # x = layers.ZeroPadding3D(padding=self.correct_pad(x, 3),
        #                         name='stem_conv_pad')(x)

        x = layers.Conv3D(self.round_filters(32), 3,
                        strides=1,
                        padding='same',
                        use_bias=False,
                        kernel_initializer=CONV_KERNEL_INITIALIZER,
                        name='stem_conv')(x)
        x = layers.BatchNormalization(axis=bn_axis, name='stem_bn')(x)
        x = layers.Activation(self.activation_fn, name='stem_activation')(x)
        blocks_args = deepcopy(DEFAULT_BLOCKS_ARGS)

        b = 0
        # Calculate the number of blocks
        blocks = float(sum(args['repeats'] for args in blocks_args))
        for (i, args) in enumerate(blocks_args):
            assert args['repeats'] > 0
            args['filters_in'] = self.round_filters(args['filters_in'])
            args['filters_out'] = self.round_filters(args['filters_out'])

            for j in range(self.round_repeats(args.pop('repeats'))):
                if j > 0:
                    args['strides'] = 1
                    args['filters_in'] = args['filters_out']
                x = self.block(x, self.activation_fn, self.drop_connect_rate * b / blocks,
                        name='block{}{}_'.format(i + 1, chr(j + 97)), **args)
                b += 1
        
        x = layers.Conv3D(self.round_filters(128), 1,
                        padding='same',
                        use_bias=False,
                        kernel_initializer=CONV_KERNEL_INITIALIZER,
                        name='top_conv')(x)
        x = layers.BatchNormalization(axis=bn_axis, name='top_bn')(x)
        x = layers.Activation(self.activation_fn, name='top_activation')(x)

        print("input.shape", inputs.shape, x.shape)
        return x

    def round_filters(self, filters):
        """Round number of filters based on depth multiplier."""
        filters *= self.width_coefficient
        divisor = self.depth_divisor
        new_filters = max(divisor, int(filters + divisor / 2) // divisor * divisor)
        # Make sure that round down does not go down by more than 10%.
        if new_filters < 0.9 * filters:
            new_filters += divisor
        return int(new_filters)

    def round_repeats(self, repeats):
        return int(math.ceil(self.depth_coefficient * repeats))

    def correct_pad(self, inputs, kernel_size):
        img_dim = 1
        input_size = backend.int_shape(inputs)[img_dim:(img_dim + 3)]


        if isinstance(kernel_size, int):
            kernel_size = (kernel_size, kernel_size, kernel_size)

        if input_size[0] is None:
            adjust = (1, 1, 1)
        else:
            adjust = (1 - input_size[0] % 2, 1 - input_size[1] % 2, 1 - input_size[2] % 2)

        correct = (kernel_size[0] // 2, kernel_size[1] // 2, kernel_size[2] // 2)

        return ((correct[0] - adjust[0], correct[0]),
                (correct[1] - adjust[1], correct[1]),
                (correct[2] - adjust[2], correct[2]))

    def block(self, inputs, activation_fn=tf.nn.swish, drop_rate=0., name='',
            filters_in=32, filters_out=16, kernel_size=3, strides=1,
            expand_ratio=1, se_ratio=0., id_skip=True):

        bn_axis = 4

        filters = filters_in * expand_ratio

        # Inverted residuals
        if expand_ratio != 1:
            x = layers.Conv3D(filters, 1,
                            padding='same',
                            use_bias=False,
                            kernel_initializer=CONV_KERNEL_INITIALIZER,
                            name=name + 'expand_conv')(inputs)
            x = layers.BatchNormalization(axis=bn_axis, name=name + 'expand_bn')(x)
            x = layers.Activation(activation_fn, name=name + 'expand_activation')(x)
        else:
            x = inputs

        # padding
        # if strides == 2:
        #     x = layers.ZeroPadding3D(padding=self.correct_pad(x, kernel_size),
        #                             name=name + 'dwconv_pad')(x)
        #     conv_pad = 'valid'
        # else:
        #     conv_pad = 'same'

        # TODO(Sicong): Find DepthwiseConv3D
        # x = layers.DepthwiseConv2D(kernel_size,
        #                         strides=strides,
        #                         padding=conv_pad,
        #                         use_bias=False,
        #                         depthwise_initializer=CONV_KERNEL_INITIALIZER,
        #                         name=name + 'dwconv')(x)
        x = layers.BatchNormalization(axis=bn_axis, name=name + 'bn')(x)
        x = layers.Activation(activation_fn, name=name + 'activation')(x)

        if 0 < se_ratio <= 1:
            filters_se = max(1, int(filters_in * se_ratio))
            se = layers.GlobalAveragePooling3D(name=name + 'se_squeeze')(x)
            se = layers.Reshape((1, 1, 1, filters), name=name + 'se_reshape')(se)
            se = layers.Conv3D(filters_se, 1,
                            padding='same',
                            activation=activation_fn,
                            kernel_initializer=CONV_KERNEL_INITIALIZER,
                            name=name + 'se_reduce')(se)
            se = layers.Conv3D(filters, 1,
                            padding='same',
                            activation='sigmoid',
                            kernel_initializer=CONV_KERNEL_INITIALIZER,
                            name=name + 'se_expand')(se)
            x = layers.multiply([x, se], name=name + 'se_excite')

        x = layers.Conv3D(filters_out, 1,
                        padding='same',
                        use_bias=False,
                        kernel_initializer=CONV_KERNEL_INITIALIZER,
                        name=name + 'project_conv')(x)
        x = layers.BatchNormalization(axis=bn_axis, name=name + 'project_bn')(x)

        if (id_skip is True and strides == 1 and filters_in == filters_out):
            if drop_rate > 0:
                x = layers.Dropout(drop_rate,
                                noise_shape=None,
                                name=name + 'drop')(x)
            x = layers.add([x, inputs], name=name + 'add')

        return x

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


if __name__ == "__main__":
    config_path = "examples/config_efficient_net.yaml"
    train(
        gpu="",
        config_path=config_path,
        gpu_allow_growth=True,
        ckpt_path="",
    )


1		"""This script provides an example of using efficient for training."""
2
3		import os
4		import math
5		import numpy as np
6		import tensorflow as tf
7		from tensorflow.keras import layers
8		from copy import deepcopy
9		from typing import List, Optional, Tuple, Union
10
11		from deepreg.model import layer
12		from deepreg.model.backbone import Backbone
13		from deepreg.model.backbone.local_net import LocalNet
14		from deepreg.model.backbone.u_net import UNet
15		from deepreg.model.layer import Extraction
16		from deepreg.registry import REGISTRY
17		from deepreg.train import train
18
19
20		EFFICIENTNET_PARAMS = {
21		# model_name: (width_mult, depth_mult, image_size, dropout_rate, dropconnect_rate)
22		"efficientnet-b0": (1.0, 1.0, 224, 0.2, 0.2),
23		"efficientnet-b1": (1.0, 1.1, 240, 0.2, 0.2),
24		"efficientnet-b2": (1.1, 1.2, 260, 0.3, 0.2),
25		"efficientnet-b3": (1.2, 1.4, 300, 0.3, 0.2),
26		"efficientnet-b4": (1.4, 1.8, 380, 0.4, 0.2),
27		"efficientnet-b5": (1.6, 2.2, 456, 0.4, 0.2),
28		"efficientnet-b6": (1.8, 2.6, 528, 0.5, 0.2),
29		"efficientnet-b7": (2.0, 3.1, 600, 0.5, 0.2),
30		}
31
32		# Each Blocks Parameters
33		DEFAULT_BLOCKS_ARGS = [
34		{'kernel_size': 3, 'repeats': 1, 'filters_in': 32, 'filters_out': 16,
35		'expand_ratio': 1, 'id_skip': True, 'strides': 1, 'se_ratio': 0.25},
36		{'kernel_size': 3, 'repeats': 2, 'filters_in': 16, 'filters_out': 24,
37		'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
38		{'kernel_size': 5, 'repeats': 2, 'filters_in': 24, 'filters_out': 40,
39		'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
40		{'kernel_size': 3, 'repeats': 3, 'filters_in': 40, 'filters_out': 80,
41		'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
42		{'kernel_size': 5, 'repeats': 3, 'filters_in': 80, 'filters_out': 112,
43		'expand_ratio': 6, 'id_skip': True, 'strides': 1, 'se_ratio': 0.25},
44		{'kernel_size': 5, 'repeats': 4, 'filters_in': 112, 'filters_out': 192,
45		'expand_ratio': 6, 'id_skip': True, 'strides': 2, 'se_ratio': 0.25},
46		{'kernel_size': 3, 'repeats': 1, 'filters_in': 192, 'filters_out': 320,
47		'expand_ratio': 6, 'id_skip': True, 'strides': 1, 'se_ratio': 0.25}
48		]
49
50		# Two Kernel Initializer
51		CONV_KERNEL_INITIALIZER = {
52		'class_name': 'VarianceScaling',
53		'config': {
54		'scale': 2.0,
55		'mode': 'fan_out',
56		'distribution': 'normal'
57		}
58		}
59
60		DENSE_KERNEL_INITIALIZER = {
61		'class_name': 'VarianceScaling',
62		'config': {
63		'scale': 1. / 3.,
64		'mode': 'fan_out',
65		'distribution': 'uniform'
66		}
67		}
68
69
70		@REGISTRY.register_backbone(name="efficient_net")
71		class EfficientNet(LocalNet):
72		"""
73		Class that implements an Efficient-Net for image registration.
74
75		Reference:
76		- Author: Mingxing Tan, Quoc V. Le,
77		EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
78		https://arxiv.org/pdf/1905.11946.pdf
79		"""
80		def __init__(
81		self,
82		image_size: tuple,
83		num_channel_initial: int,
84		extract_levels: Tuple[int, ...],
85		out_kernel_initializer: str,
86		out_activation: str,
87		out_channels: int,
88		depth: Optional[int] = None,
89		pooling: bool = True,
90		concat_skip: bool = False,
91		width_coefficient: float = 1.0,
92		depth_coefficient: float = 1.0,
93		default_size: int = 224,
94		dropout_rate: float = 0.2,
95		drop_connect_rate: float = 0.2,
96		depth_divisor: int = 8,
97		name: str = "EfficientNet",
98		**kwargs,
99		):
100		"""
101		Init.
102
103		Image is encoded gradually, i from level 0 to D,
104		then it is decoded gradually, j from level D to 0.
105		Some of the decoded levels are used for generating extractions.
106
107		So, extract_levels are between [0, D].
108
109		:param image_size: such as (dim1, dim2, dim3)
110		:param num_channel_initial: number of initial channels.
111		:param extract_levels: from which depths the output will be built.
112		:param out_kernel_initializer: initializer to use for kernels.
113		:param out_activation: activation to use at end layer.
114		:param out_channels: number of channels for the extractions
115		:param depth: depth of the encoder.
116		If depth is not given, depth = max(extract_levels) will be used.
117		:param pooling: for down-sampling, use non-parameterized
118		pooling if true, otherwise use conv3d
119		:param concat_skip: when up-sampling, concatenate skipped
120		tensor if true, otherwise use addition
121		:param width_coefficient: float, scaling coefficient for network width.
122		:param depth_coefficient: float, scaling coefficient for network depth.
123		:param default_size: int, default input image size.
124		:param dropout_rate: float, dropout rate before final classifier layer.
125		:param drop_connect_rate: float, dropout rate at skip connections.
126		:param depth_divisor: int divisor for depth.
127		:param name: name of the backbone.
128		:param kwargs: additional arguments.
129		"""
130		if depth is None:
131		depth = max(extract_levels)
132		kwargs["encode_kernel_sizes"] = [7] + [3] * depth
133		super().__init__(
134		image_size=image_size,
135		num_channel_initial=num_channel_initial,
136		depth=depth,
137		extract_levels=extract_levels,
138		out_kernel_initializer=out_kernel_initializer,
139		out_activation=out_activation,
140		out_channels=out_channels,
141		use_additive_upsampling = False,
142		pooling=pooling,
143		concat_skip=concat_skip,
144		name=name,
145		**kwargs,
146		)
147
148		self.width_coefficient = width_coefficient
149		self.depth_coefficient = depth_coefficient
150		self.default_size = default_size
151		self.dropout_rate = dropout_rate
152		self.drop_connect_rate = drop_connect_rate
153		self.depth_divisor = depth_divisor
154		self.activation_fn = tf.nn.swish
155
156	View Code Duplication	def call(self, inputs: tf.Tensor, training=None, mask=None) -> tf.Tensor:
		0 ignored issues – show Duplication introduced 2021-06-11 15:30 UTC by Report Bug Copy Issue Report This code seems to be duplicated in your project. Loading history...
157		"""
158		Build compute graph based on built layers.
159
160		:param inputs: image batch, shape = (batch, f_dim1, f_dim2, f_dim3, ch)
161		:param training: None or bool.
162		:param mask: None or tf.Tensor.
163		:return: shape = (batch, f_dim1, f_dim2, f_dim3, out_channels)
164		"""
165
166		# encoding / down-sampling
167		skips = []
168		encoded = inputs
169		for d in range(self._depth):
170		skip = self._encode_convs[d](inputs=encoded, training=training)
171		encoded = self._encode_pools[d](inputs=skip, training=training)
172		skips.append(skip)
173
174		# bottom
175		decoded = self.build_efficient_net(inputs=encoded, training=training) # type: ignore
176
177		# decoding / up-sampling
178		outs = [decoded]
179		for d in range(self._depth - 1, min(self._extract_levels) - 1, -1):
180		decoded = self._decode_deconvs[d](inputs=decoded, training=training)
181		decoded = self.build_skip_block()([decoded, skips[d]])
182		decoded = self._decode_convs[d](inputs=decoded, training=training)
183		outs = [decoded] + outs
184
185		# output
186		output = self._output_block(outs) # type: ignore
187
188		return output
189
190
191		def build_efficient_net(self, inputs: tf.Tensor, training=None) -> tf.Tensor:
192		"""
193		Builds graph based on built layers.
194
195		:param inputs: shape = (batch, f_dim1, f_dim2, f_dim3, in_channels)
196		:param training:
197		:param mask:
198		:return: shape = (batch, f_dim1, f_dim2, f_dim3, out_channels)
199		"""
200		img_input = layers.Input(tensor=inputs, shape=self.image_size)
201		bn_axis = 4
202		x = img_input
203		# x = layers.ZeroPadding3D(padding=self.correct_pad(x, 3),
204		# name='stem_conv_pad')(x)
205
206		x = layers.Conv3D(self.round_filters(32), 3,
207		strides=1,
208		padding='same',
209		use_bias=False,
210		kernel_initializer=CONV_KERNEL_INITIALIZER,
211		name='stem_conv')(x)
212		x = layers.BatchNormalization(axis=bn_axis, name='stem_bn')(x)
213		x = layers.Activation(self.activation_fn, name='stem_activation')(x)
214		blocks_args = deepcopy(DEFAULT_BLOCKS_ARGS)
215
216		b = 0
217		# Calculate the number of blocks
218		blocks = float(sum(args['repeats'] for args in blocks_args))
219		for (i, args) in enumerate(blocks_args):
220		assert args['repeats'] > 0
221		args['filters_in'] = self.round_filters(args['filters_in'])
222		args['filters_out'] = self.round_filters(args['filters_out'])
223
224		for j in range(self.round_repeats(args.pop('repeats'))):
225		if j > 0:
226		args['strides'] = 1
227		args['filters_in'] = args['filters_out']
228		x = self.block(x, self.activation_fn, self.drop_connect_rate * b / blocks,
229		name='block{}{}_'.format(i + 1, chr(j + 97)), **args)
230		b += 1
231
232		x = layers.Conv3D(self.round_filters(128), 1,
233		padding='same',
234		use_bias=False,
235		kernel_initializer=CONV_KERNEL_INITIALIZER,
236		name='top_conv')(x)
237		x = layers.BatchNormalization(axis=bn_axis, name='top_bn')(x)
238		x = layers.Activation(self.activation_fn, name='top_activation')(x)
239
240		print("input.shape", inputs.shape, x.shape)
241		return x
242
243		def round_filters(self, filters):
244		"""Round number of filters based on depth multiplier."""
245		filters *= self.width_coefficient
246		divisor = self.depth_divisor
247		new_filters = max(divisor, int(filters + divisor / 2) // divisor * divisor)
248		# Make sure that round down does not go down by more than 10%.
249		if new_filters < 0.9 * filters:
250		new_filters += divisor
251		return int(new_filters)
252
253		def round_repeats(self, repeats):
254		return int(math.ceil(self.depth_coefficient * repeats))
255
256		def correct_pad(self, inputs, kernel_size):
257		img_dim = 1
258		input_size = backend.int_shape(inputs)[img_dim:(img_dim + 3)]
		0 ignored issues – show Comprehensibility Best Practice introduced 2021-06-11 15:30 UTC by Report Bug Copy Issue Report The variable `backend` does not seem to be defined. Loading history...
259
260		if isinstance(kernel_size, int):
261		kernel_size = (kernel_size, kernel_size, kernel_size)
262
263		if input_size[0] is None:
264		adjust = (1, 1, 1)
265		else:
266		adjust = (1 - input_size[0] % 2, 1 - input_size[1] % 2, 1 - input_size[2] % 2)
267
268		correct = (kernel_size[0] // 2, kernel_size[1] // 2, kernel_size[2] // 2)
269
270		return ((correct[0] - adjust[0], correct[0]),
271		(correct[1] - adjust[1], correct[1]),
272		(correct[2] - adjust[2], correct[2]))
273
274		def block(self, inputs, activation_fn=tf.nn.swish, drop_rate=0., name='',
275		filters_in=32, filters_out=16, kernel_size=3, strides=1,
276		expand_ratio=1, se_ratio=0., id_skip=True):
277
278		bn_axis = 4
279
280		filters = filters_in * expand_ratio
281
282		# Inverted residuals
283		if expand_ratio != 1:
284		x = layers.Conv3D(filters, 1,
285		padding='same',
286		use_bias=False,
287		kernel_initializer=CONV_KERNEL_INITIALIZER,
288		name=name + 'expand_conv')(inputs)
289		x = layers.BatchNormalization(axis=bn_axis, name=name + 'expand_bn')(x)
290		x = layers.Activation(activation_fn, name=name + 'expand_activation')(x)
291		else:
292		x = inputs
293
294		# padding
295		# if strides == 2:
296		# x = layers.ZeroPadding3D(padding=self.correct_pad(x, kernel_size),
297		# name=name + 'dwconv_pad')(x)
298		# conv_pad = 'valid'
299		# else:
300		# conv_pad = 'same'
301
302		# TODO(Sicong): Find DepthwiseConv3D
303		# x = layers.DepthwiseConv2D(kernel_size,
304		# strides=strides,
305		# padding=conv_pad,
306		# use_bias=False,
307		# depthwise_initializer=CONV_KERNEL_INITIALIZER,
308		# name=name + 'dwconv')(x)
309		x = layers.BatchNormalization(axis=bn_axis, name=name + 'bn')(x)
310		x = layers.Activation(activation_fn, name=name + 'activation')(x)
311
312		if 0 < se_ratio <= 1:
313		filters_se = max(1, int(filters_in * se_ratio))
314		se = layers.GlobalAveragePooling3D(name=name + 'se_squeeze')(x)
315		se = layers.Reshape((1, 1, 1, filters), name=name + 'se_reshape')(se)
316		se = layers.Conv3D(filters_se, 1,
317		padding='same',
318		activation=activation_fn,
319		kernel_initializer=CONV_KERNEL_INITIALIZER,
320		name=name + 'se_reduce')(se)
321		se = layers.Conv3D(filters, 1,
322		padding='same',
323		activation='sigmoid',
324		kernel_initializer=CONV_KERNEL_INITIALIZER,
325		name=name + 'se_expand')(se)
326		x = layers.multiply([x, se], name=name + 'se_excite')
327
328		x = layers.Conv3D(filters_out, 1,
329		padding='same',
330		use_bias=False,
331		kernel_initializer=CONV_KERNEL_INITIALIZER,
332		name=name + 'project_conv')(x)
333		x = layers.BatchNormalization(axis=bn_axis, name=name + 'project_bn')(x)
334
335		if (id_skip is True and strides == 1 and filters_in == filters_out):
336		if drop_rate > 0:
337		x = layers.Dropout(drop_rate,
338		noise_shape=None,
339		name=name + 'drop')(x)
340		x = layers.add([x, inputs], name=name + 'add')
341
342		return x
343
344		def get_config(self) -> dict:
345		"""Return the config dictionary for recreating this class."""
346		config = super().get_config()
347		return config
348
349
350		if __name__ == "__main__":
351		config_path = "examples/config_efficient_net.yaml"
352		train(
353		gpu="",
354		config_path=config_path,
355		gpu_allow_growth=True,
356		ckpt_path="",
357		)
358

DeepRegNet / DeepReg

Pull Request — main (#785)

efficient_net.EfficientNet.build_efficient_net() A

Complexity

Size

Duplication

Importance

How to fix Long Method

Long Method

Duplication Side-by-Side

Filter issues like