def test_normalize_data_format():
assert keras_utils.normalize_data_format(
"Channels_Last") == "channels_last"
assert keras_utils.normalize_data_format(
"CHANNELS_FIRST") == "channels_first"
with pytest.raises(ValueError,
match="The `data_format` argument must be one of"):
keras_utils.normalize_data_format("invalid")
def __init__(self,
reduce_function: Callable,
output_size: Union[int, Iterable[int]],
data_format=None,
**kwargs):
self.data_format = conv_utils.normalize_data_format(data_format)
self.reduce_function = reduce_function
self.output_size = conv_utils.normalize_tuple(output_size, 1,
"output_size")
super().__init__(**kwargs)
def __init__(self,
bins: Union[Iterable[int], Iterable[Iterable[int]]],
data_format=None,
*args,
**kwargs):
self.bins = [conv_utils.normalize_tuple(bin, 2, "bin") for bin in bins]
self.data_format = conv_utils.normalize_data_format(data_format)
self.pool_layers = []
for bin in self.bins:
self.pool_layers.append(
AdaptiveAveragePooling2D(bin, self.data_format))
super().__init__(*args, **kwargs)
def _norm_params(images, mask_size, data_format):
mask_size = tf.convert_to_tensor(mask_size)
if tf.executing_eagerly():
tf.assert_equal(
tf.reduce_any(mask_size % 2 != 0),
False,
"mask_size should be divisible by 2",
)
if tf.rank(mask_size) == 0:
mask_size = tf.stack([mask_size, mask_size])
data_format = keras_utils.normalize_data_format(data_format)
image_height, image_width = _get_image_wh(images, data_format)
return mask_size, data_format, image_height, image_width
