One-liner APIs

tf_keras_vis/activation_maximization/input_modifiers.py

@@ -5,6 +5,8 @@
 import tensorflow as tf
 from scipy.ndimage.interpolation import rotate, zoom
 
+from ..utils import order
+
 
 class InputModifier(ABC):
  """Abstract class for defining an input modifier.
@@ -51,12 +53,15 @@ def __call__(self, seed_input) -> np.ndarray:
 class Rotate(InputModifier):
  """An input modifier that introduces random rotation.
  """
- def __init__(self, axes=(1, 2), degree=3.0) -> None:
+ def __init__(self, axes=(1, 2), degree=3.0, interpolation='bilinear') -> None:
  """
  Args:
  axes: The two axes that define the plane of rotation.
  Defaults to (1, 2).
  degree: The amount of rotation to apply. Defaults to 3.0.
+ interpolation: An integer or string. When integer, `interpolation`'s specification is
+ the same as `order` option of scipy-ndimage API. When string, `interpolation` MUST
+ be one of `"nearest"`, `"bilinear"` and `"cubic"`. Defaults to `"bilinear"`.
 
  Raises:
  ValueError: When axes is not a tuple of two ints.
@@ -68,6 +73,7 @@ def __init__(self, axes=(1, 2), degree=3.0) -> None:
  self.axes = axes
  self.degree = float(degree)
  self.random_generator = np.random.default_rng()
+ self.order = order(interpolation)
 
  def __call__(self, seed_input) -> np.ndarray:
  ndim = len(seed_input.shape)
@@ -80,7 +86,7 @@ def __call__(self, seed_input) -> np.ndarray:
  self.random_generator.uniform(-self.degree, self.degree),
  axes=self.axes,
  reshape=False,
- order=1,
+ order=self.order,
  mode='reflect',
  prefilter=False)
  return seed_input
@@ -89,26 +95,33 @@ def __call__(self, seed_input) -> np.ndarray:
 class Rotate2D(Rotate):
  """An input modifier for 2D that introduces random rotation.
  """
- def __init__(self, degree=3.0) -> None:
+ def __init__(self, degree=3.0, interpolation='bilinear') -> None:
  """
  Args:
  degree: The amount of rotation to apply. Defaults to 3.0.
+ interpolation: An integer or string. When integer, `interpolation`'s specification is
+ the same as `order` option of scipy-ndimage API. When string, `interpolation` MUST
+ be one of `"nearest"`, `"bilinear"` and `"cubic"`. Defaults to `"bilinear"`.
  """
- super().__init__(axes=(1, 2), degree=degree)
+ super().__init__(axes=(1, 2), degree=degree, interpolation=interpolation)
 
 
 class Scale(InputModifier):
  """An input modifier that introduces randam scaling.
  """
- def __init__(self, low=0.9, high=1.1) -> None:
+ def __init__(self, low=0.9, high=1.1, interpolation='bilinear') -> None:
  """
  Args:
  low (float, optional): Lower boundary of the zoom factor. Defaults to 0.9.
  high (float, optional): Higher boundary of the zoom factor. Defaults to 1.1.
+ interpolation: An integer or string. When integer, `interpolation`'s specification is
+ the same as `order` option of scipy-ndimage API. When string, `interpolation` MUST
+ be one of `"nearest"`, `"bilinear"` and `"cubic"`. Defaults to `"bilinear"`.
  """
  self.low = low
  self.high = high
  self.random_generator = np.random.default_rng()
+ self.order = order(interpolation)
 
  def __call__(self, seed_input) -> np.ndarray:
  ndim = len(seed_input.shape)
@@ -121,7 +134,7 @@ def __call__(self, seed_input) -> np.ndarray:
  _factor = factor = self.random_generator.uniform(self.low, self.high)
  factor *= np.ones(ndim - 2)
  factor = (1, ) + tuple(factor) + (1, )
- seed_input = zoom(seed_input, factor, order=1, mode='reflect', prefilter=False)
+ seed_input = zoom(seed_input, factor, order=self.order, mode='reflect', prefilter=False)
  if _factor > 1.0:
  indices = (self._central_crop_range(x, e) for x, e in zip(seed_input.shape, shape))
  indices = (slice(start, stop) for start, stop in indices)

tf_keras_vis/utils/__init__.py

@@ -1,4 +1,5 @@
 import os
+from collections.abc import Iterable
 from typing import Tuple
 
 import numpy as np
@@ -39,27 +40,38 @@ def num_of_gpus() -> Tuple[int, int]:
  return 0, 0
 
 
-def listify(value, return_empty_list_if_none=True, convert_tuple_to_list=True) -> list:
- """Ensures that the value is a list.
+def listify(value,
+ return_empty_list_if_none=True,
+ convert_tuple_to_list=True,
+ convert_iterable_to_list=False) -> list:
+ """Ensures that `value` is a list.
 
- If it is not a list, it creates a new list with `value` as an item.
+ If `value` is not a list, this function creates an new list that includes `value`.
 
  Args:
  value (object): A list or something else.
- return_empty_list_if_none (bool, optional): When True (default), None you passed as `value`
- will be converted to a empty list (i.e., `[]`). When False, None will be converted to
- a list that has an None (i.e., `[None]`). Defaults to True.
- convert_tuple_to_list (bool, optional): When True (default), a tuple you passed as `value`
- will be converted to a list. When False, a tuple will be unconverted
- (i.e., returning a tuple object that was passed as `value`). Defaults to True.
+ return_empty_list_if_none (bool, optional): When True (default), `None` you passed as
+ `value` will be converted to a empty list (i.e., `[]`). When False, `None` will be
+ converted to a list that contains an `None` (i.e., `[None]`). Defaults to True.
+ convert_tuple_to_list (bool, optional):When True (default), a tuple object you
+ passed as `value` will be converted to a list. When False, a tuple object will be
+ unconverted (i.e., returning a list of a tuple object). Defaults to True.
+ convert_iterable_to_list (bool, optional): When True (default), an iterable object you
+ passed as `value` will be converted to a list. When False, an iterable object will be
+ unconverted (i.e., returning a list of an iterable object). Defaults to False.
  Returns:
- list: A list. When `value` is a tuple and `convert_tuple_to_list` is False, a tuple.
+ list: A list
  """
  if not isinstance(value, list):
  if value is None and return_empty_list_if_none:
  value = []
  elif isinstance(value, tuple) and convert_tuple_to_list:
  value = list(value)
+ elif isinstance(value, Iterable) and convert_iterable_to_list:
+ if not convert_tuple_to_list:
+ raise ValueError("When 'convert_tuple_to_list' option is False,"
+ "'convert_iterable_to_list' option should also be False.")
+ value = list(value)
  else:
  value = [value]
  return value
@@ -132,3 +144,18 @@ def lower_precision_dtype(model):
  (isinstance(layer, tf.keras.Model) and is_mixed_precision(layer)):
  return layer.compute_dtype
  return model.dtype # pragma: no cover
+
+
+def order(value):
+ if isinstance(value, int):
+ return value
+ if isinstance(value, str):
+ value = value.lower()
+ if value == 'nearest':
+ return 0
+ if value == 'bilinear':
+ return 1
+ if value == 'cubic':
+ return 3
+ raise ValueError(f"{value} is not supported. "
+ "The value MUST be an integer or one of 'nearest', 'bilinear' or 'cubic'.")