本文整理匯總了Python中inception_preprocessing.distort_color方法的典型用法代碼示例。如果您正苦於以下問題：Python inception_preprocessing.distort_color方法的具體用法？Python inception_preprocessing.distort_color怎麽用？Python inception_preprocessing.distort_color使用的例子？那麽, 這裏精選的方法代碼示例或許可以為您提供幫助。您也可以進一步了解該方法所在類inception_preprocessing的用法示例。
在下文中一共展示了inception_preprocessing.distort_color方法的2個代碼示例，這些例子默認根據受歡迎程度排序。您可以為喜歡或者感覺有用的代碼點讚，您的評價將有助於係統推薦出更棒的Python代碼示例。
示例1: augment_image
# 需要導入模塊: import inception_preprocessing [as 別名]
# 或者: from inception_preprocessing import distort_color [as 別名]
def augment_image(image):
"""Augmentation the image with a random modification.
Args:
image: input Tensor image of rank 3, with the last dimension
of size 3.
Returns:
Distorted Tensor image of the same shape.
"""
with tf.variable_scope('AugmentImage'):
height = image.get_shape().dims[0].value
width = image.get_shape().dims[1].value
# Random crop cut from the street sign image, resized to the same size.
# Assures that the crop is covers at least 0.8 area of the input image.
bbox_begin, bbox_size, _ = tf.image.sample_distorted_bounding_box(
tf.shape(image),
bounding_boxes=tf.zeros([0, 0, 4]),
min_object_covered=0.8,
aspect_ratio_range=[0.8, 1.2],
area_range=[0.8, 1.0],
use_image_if_no_bounding_boxes=True)
distorted_image = tf.slice(image, bbox_begin, bbox_size)
# Randomly chooses one of the 4 interpolation methods
distorted_image = inception_preprocessing.apply_with_random_selector(
distorted_image,
lambda x, method: tf.image.resize_images(x, [height, width], method),
num_cases=4)
distorted_image.set_shape([height, width, 3])
# Color distortion
distorted_image = inception_preprocessing.apply_with_random_selector(
distorted_image,
functools.partial(
inception_preprocessing.distort_color, fast_mode=False),
num_cases=4)
distorted_image = tf.clip_by_value(distorted_image, -1.5, 1.5)
return distorted_image 示例2: augment_image
# 需要導入模塊: import inception_preprocessing [as 別名]
# 或者: from inception_preprocessing import distort_color [as 別名]
def augment_image(image):
"""Augmentation the image with a random modification.
Args:
image: input Tensor image of rank 3, with the last dimension
of size 3.
Returns:
Distorted Tensor image of the same shape.
"""
with tf.variable_scope('AugmentImage'):
height = image.get_shape().dims[0].value
width = image.get_shape().dims[1].value
# Random crop cut from the street sign image, resized to the same size.
# Assures that the crop is covers at least 0.8 area of the input image.
bbox_begin, bbox_size, _ = tf.image.sample_distorted_bounding_box(
tf.shape(image),
bounding_boxes=tf.zeros([0, 0, 4]),
min_object_covered=0.8,
aspect_ratio_range=[0.8, 1.2],
area_range=[0.8, 1.0],
use_image_if_no_bounding_boxes=True)
distorted_image = tf.slice(image, bbox_begin, bbox_size)
# Randomly chooses one of the 4 interpolation methods
distorted_image = inception_preprocessing.apply_with_random_selector(
distorted_image,
lambda x, method: tf.image.resize_images(x, [height, width], method),
num_cases=4)
distorted_image.set_shape([height, width, 3])
# Color distortion
# TODO:incompatible with clip value in inception_preprocessing.distort_color
distorted_image = inception_preprocessing.apply_with_random_selector(
distorted_image,
functools.partial(
inception_preprocessing.distort_color, fast_mode=False),
num_cases=4)
distorted_image = tf.clip_by_value(distorted_image, -1.5, 1.5)
return distorted_image