custom_image_label_loss.RootMeanSquaredDifference.call() - Code Metrics - DeepRegNet/DeepReg - Measure and Improve Code Quality continuously with Scrutinizer

RootMeanSquaredDifference.call() A
last analyzed 2022-05-18 21:56 UTC

↳ Parent: custom_image_label_loss

Complexity

Conditions

Size

Total Lines	13
Code Lines	6

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
eloc	6
dl	0
loc	13
rs	10
c	0
b	0
f	0
cc	1
nop	3

"""This script provides an example of using custom backbone for training."""
import tensorflow as tf

from deepreg.registry import REGISTRY
from deepreg.train import train


@REGISTRY.register_loss(name="root_mean_square")
class RootMeanSquaredDifference(tf.keras.losses.Loss):
    """
    Square root of the mean of squared distance between y_true and y_pred.

    y_true and y_pred have to be at least 1d tensor, including batch axis.
    """

    def __init__(
        self,
        name: str = "RootMeanSquaredDifference",
        **kwargs,
    ):
        """
        Init.

        :param name: name of the loss
        :param kwargs: additional arguments.
        """
        super().__init__(name=name, **kwargs)
        self.flatten = tf.keras.layers.Flatten()

    def call(self, y_true: tf.Tensor, y_pred: tf.Tensor) -> tf.Tensor:
        """
        Return loss for a batch.

        :param y_true: shape = (batch, ...)
        :param y_pred: shape = (batch, ...)
        :return: shape = (batch,)
        """
        loss = tf.math.squared_difference(y_true, y_pred)
        loss = self.flatten(loss)
        loss = tf.reduce_mean(loss, axis=1)
        loss = tf.math.sqrt(loss)
        return loss


config_path = "examples/config_custom_image_label_loss.yaml"
train(
    gpu="",
    config_path=config_path,
    gpu_allow_growth=True,
    ckpt_path="",
)


1			"""This script provides an example of using custom backbone for training."""
2			import tensorflow as tf
3
4			from deepreg.registry import REGISTRY
5			from deepreg.train import train
6
7
8			@REGISTRY.register_loss(name="root_mean_square")
9			class RootMeanSquaredDifference(tf.keras.losses.Loss):
10			"""
11			Square root of the mean of squared distance between y_true and y_pred.
12
13			y_true and y_pred have to be at least 1d tensor, including batch axis.
14			"""
15
16			def __init__(
17			self,
18			name: str = "RootMeanSquaredDifference",
19			**kwargs,
20			):
21			"""
22			Init.
23
24			:param name: name of the loss
25			:param kwargs: additional arguments.
26			"""
27			super().__init__(name=name, **kwargs)
28			self.flatten = tf.keras.layers.Flatten()
29
30			def call(self, y_true: tf.Tensor, y_pred: tf.Tensor) -> tf.Tensor:
31			"""
32			Return loss for a batch.
33
34			:param y_true: shape = (batch, ...)
35			:param y_pred: shape = (batch, ...)
36			:return: shape = (batch,)
37			"""
38			loss = tf.math.squared_difference(y_true, y_pred)
39			loss = self.flatten(loss)
40			loss = tf.reduce_mean(loss, axis=1)
41			loss = tf.math.sqrt(loss)
42			return loss
43
44
45			config_path = "examples/config_custom_image_label_loss.yaml"
46			train(
47			gpu="",
48			config_path=config_path,
49			gpu_allow_growth=True,
50			ckpt_path="",
51			)
52

DeepRegNet / DeepReg

RootMeanSquaredDifference.call() A last analyzed 2022-05-18 21:56 UTC

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like

RootMeanSquaredDifference.call() A
last analyzed 2022-05-18 21:56 UTC