def new_method(self, *args, **kwargs):
[threshold], _ = parse_user_args(method, *args, **kwargs)
type_check(threshold, (tuple, ), "threshold")
type_check_list(threshold, (int, ), "threshold")
if len(threshold) != 2:
raise ValueError("threshold must be a sequence of two numbers.")
for element in threshold:
check_value(element, (0, UINT8_MAX))
if threshold[1] < threshold[0]:
raise ValueError("threshold must be in min max format numbers.")
return method(self, *args, **kwargs)
def check_crop_size(size):
"""Wrapper method to check the parameters of crop size."""
type_check(size, (int, list, tuple), "size")
if isinstance(size, int):
check_value(size, (1, FLOAT_MAX_INTEGER))
elif isinstance(size, (tuple, list)) and len(size) == 2:
for index, value in enumerate(size):
type_check(value, (int, ), "size[{}]".format(index))
check_value(value, (1, FLOAT_MAX_INTEGER))
else:
raise TypeError(
"Size should be a single integer or a list/tuple (h, w) of length 2."
)
def check_random_color_adjust_param(value, input_name, center=1, bound=(0, FLOAT_MAX_INTEGER), non_negative=True):
"""Check the parameters in random color adjust operation."""
type_check(value, (numbers.Number, list, tuple), input_name)
if isinstance(value, numbers.Number):
if value < 0:
raise ValueError("The input value of {} cannot be negative.".format(input_name))
elif isinstance(value, (list, tuple)):
if len(value) != 2:
raise TypeError("If {0} is a sequence, the length must be 2.".format(input_name))
if value[0] > value[1]:
raise ValueError("{0} value should be in (min,max) format. Got ({1}, {2}).".format(input_name,
value[0], value[1]))
check_range(value, bound)
def new_method(self, *args, **kwargs):
[transformation_matrix, mean_vector], _ = parse_user_args(method, *args, **kwargs)
type_check(transformation_matrix, (np.ndarray,), "transformation_matrix")
type_check(mean_vector, (np.ndarray,), "mean_vector")
if transformation_matrix.shape[0] != transformation_matrix.shape[1]:
raise ValueError("transformation_matrix should be a square matrix. "
"Got shape {} instead".format(transformation_matrix.shape))
if mean_vector.shape[0] != transformation_matrix.shape[0]:
raise ValueError("mean_vector length {0} should match either one dimension of the square"
"transformation_matrix {1}.".format(mean_vector.shape[0], transformation_matrix.shape))
return method(self, *args, **kwargs)
def check_degrees(degrees):
"""Check if the degrees is legal."""
type_check(degrees, (numbers.Number, list, tuple), "degrees")
if isinstance(degrees, numbers.Number):
check_value(degrees, (0, float("inf")), "degrees")
elif isinstance(degrees, (list, tuple)):
if len(degrees) == 2:
type_check_list(degrees, (numbers.Number, ), "degrees")
if degrees[0] > degrees[1]:
raise ValueError(
"degrees should be in (min,max) format. Got (max,min).")
else:
raise TypeError("If degrees is a sequence, the length must be 2.")
def new_method(self, *args, **kwargs):
[size, padding, pad_if_needed, fill_value,
padding_mode], _ = parse_user_args(method, *args, **kwargs)
check_crop_size(size)
type_check(pad_if_needed, (bool, ), "pad_if_needed")
if padding is not None:
check_padding(padding)
if fill_value is not None:
check_fill_value(fill_value)
if padding_mode is not None:
type_check(padding_mode, (Border, ), "padding_mode")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[kernel_size, sigma], _ = parse_user_args(method, *args, **kwargs)
type_check(kernel_size, (int, list, tuple), "kernel_size")
if isinstance(kernel_size, int):
check_value(kernel_size, (1, FLOAT_MAX_INTEGER), "kernel_size")
check_odd(kernel_size, "kernel_size")
elif isinstance(kernel_size, (list, tuple)) and len(kernel_size) == 2:
for index, value in enumerate(kernel_size):
type_check(value, (int, ), "kernel_size[{}]".format(index))
check_value(value, (1, FLOAT_MAX_INTEGER), "kernel_size")
check_odd(value, "kernel_size[{}]".format(index))
else:
raise TypeError(
"Kernel size should be a single integer or a list/tuple (kernel_width, kernel_height) of length 2."
)
if sigma is not None:
type_check(sigma, (numbers.Number, list, tuple), "sigma")
if isinstance(sigma, numbers.Number):
check_value(sigma, (0, FLOAT_MAX_INTEGER), "sigma")
elif isinstance(sigma, (list, tuple)) and len(sigma) == 2:
for index, value in enumerate(sigma):
type_check(value, (numbers.Number, ),
"size[{}]".format(index))
check_value(value, (0, FLOAT_MAX_INTEGER), "sigma")
else:
raise TypeError(
"Sigma should be a single number or a list/tuple of length 2 for width and height."
)
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[bits], _ = parse_user_args(method, *args, **kwargs)
if bits is not None:
type_check(bits, (list, tuple, int), "bits")
if isinstance(bits, int):
check_value(bits, [1, 8])
if isinstance(bits, (list, tuple)):
if len(bits) != 2:
raise TypeError("Size of bits should be a single integer or a list/tuple (min, max) of length 2.")
for item in bits:
check_uint8(item, "bits")
# also checks if min <= max
check_range(bits, [1, 8])
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[degrees, resample, expand, center, fill_value], _ = parse_user_args(method, *args, **kwargs)
check_degrees(degrees)
if resample is not None:
type_check(resample, (Inter,), "resample")
if expand is not None:
type_check(expand, (bool,), "expand")
if center is not None:
check_2tuple(center, "center")
if fill_value is not None:
check_fill_value(fill_value)
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[policy], _ = parse_user_args(method, *args, **kwargs)
type_check(policy, (list,), "policy")
if not policy:
raise ValueError("policy can not be empty.")
for sub_ind, sub in enumerate(policy):
type_check(sub, (list,), "policy[{0}]".format([sub_ind]))
if not sub:
raise ValueError("policy[{0}] can not be empty.".format(sub_ind))
for op_ind, tp in enumerate(sub):
check_2tuple(tp, "policy[{0}][{1}]".format(sub_ind, op_ind))
check_tensor_op(tp[0], "op of (op, prob) in policy[{0}][{1}]".format(sub_ind, op_ind))
check_value(tp[1], (0, 1), "prob of (op, prob) policy[{0}][{1}]".format(sub_ind, op_ind))
return method(self, *args, **kwargs)
def check_random_color_adjust_param(value,
input_name,
center=1,
bound=(0, FLOAT_MAX_INTEGER),
non_negative=True):
"""Check the parameters in random color adjust operation."""
type_check(value, (numbers.Number, list, tuple), input_name)
if isinstance(value, numbers.Number):
if value < 0:
raise ValueError(
"The input value of {} cannot be negative.".format(input_name))
elif isinstance(value, (list, tuple)) and len(value) == 2:
check_range(value, bound)
if value[0] > value[1]:
raise ValueError(
"value should be in (min,max) format. Got (max,min).")
def new_method(self, *args, **kwargs):
[transforms, num_ops], _ = parse_user_args(method, *args, **kwargs)
type_check(num_ops, (int, ), "num_ops")
check_positive(num_ops, "num_ops")
if num_ops > len(transforms):
raise ValueError("num_ops is greater than transforms list size.")
parsed_transforms = []
for op in transforms:
if op and getattr(op, 'parse', None):
parsed_transforms.append(op.parse())
else:
parsed_transforms.append(op)
type_check_list(parsed_transforms, (TensorOp, TensorOperation),
"transforms")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[transforms, num_ops], _ = parse_user_args(method, *args, **kwargs)
type_check(transforms, (list,), "transforms")
if not transforms:
raise ValueError("transforms list is empty.")
for transform in transforms:
if isinstance(transform, TensorOp):
raise ValueError("transform list only accepts Python operations.")
type_check(num_ops, (int,), "num_ops")
check_positive(num_ops, "num_ops")
if num_ops > len(transforms):
raise ValueError("num_ops cannot be greater than the length of transforms list.")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[prob, scale, ratio, value, inplace,
max_attempts], _ = parse_user_args(method, *args, **kwargs)
type_check(prob, (
float,
int,
), "prob")
type_check_list(scale, (
float,
int,
), "scale")
if len(scale) != 2:
raise TypeError("scale should be a list or tuple of length 2.")
type_check_list(ratio, (
float,
int,
), "ratio")
if len(ratio) != 2:
raise TypeError("ratio should be a list or tuple of length 2.")
type_check(value, (int, list, tuple, str), "value")
type_check(inplace, (bool, ), "inplace")
type_check(max_attempts, (int, ), "max_attempts")
check_erasing_value(value)
check_value(prob, [0., 1.], "prob")
if scale[0] > scale[1]:
raise ValueError(
"scale should be in (min,max) format. Got (max,min).")
check_range(scale, [0, FLOAT_MAX_INTEGER])
check_positive(scale[1], "scale[1]")
if ratio[0] > ratio[1]:
raise ValueError(
"ratio should be in (min,max) format. Got (max,min).")
check_value_ratio(ratio[0], [0, FLOAT_MAX_INTEGER])
check_value_ratio(ratio[1], [0, FLOAT_MAX_INTEGER])
if isinstance(value, int):
check_value(value, (0, 255))
if isinstance(value, (list, tuple)):
for item in value:
type_check(item, (int, ), "value")
check_value(item, [0, 255], "value")
check_value(max_attempts, (1, FLOAT_MAX_INTEGER))
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[prob, scale, ratio, value, inplace, max_attempts], _ = parse_user_args(method, *args, **kwargs)
check_value(prob, [0., 1.], "prob")
if scale[0] > scale[1]:
raise ValueError("scale should be in (min,max) format. Got (max,min).")
check_range(scale, [0, FLOAT_MAX_INTEGER])
check_positive(scale[1], "scale[1]")
if ratio[0] > ratio[1]:
raise ValueError("ratio should be in (min,max) format. Got (max,min).")
check_range(ratio, [0, FLOAT_MAX_INTEGER])
check_positive(ratio[0], "ratio[0]")
check_positive(ratio[1], "ratio[1]")
check_erasing_value(value)
type_check(inplace, (bool,), "inplace")
check_value(max_attempts, (1, FLOAT_MAX_INTEGER))
return method(self, *args, **kwargs)
def check_size_scale_ration_max_attempts_paras(size, scale, ratio, max_attempts):
"""Wrapper method to check the parameters of RandomCropDecodeResize and SoftDvppDecodeRandomCropResizeJpeg."""
check_crop_size(size)
if scale is not None:
type_check(scale, (tuple,), "scale")
type_check_list(scale, (float, int), "scale")
check_range(scale, [0, FLOAT_MAX_INTEGER])
if scale[0] > scale[1]:
raise ValueError("scale should be in (min,max) format. Got (max,min).")
if ratio is not None:
type_check(ratio, (tuple,), "ratio")
type_check_list(ratio, (float, int), "ratio")
check_range(ratio, [0, FLOAT_MAX_INTEGER])
if ratio[0] > ratio[1]:
raise ValueError("ratio should be in (min,max) format. Got (max,min).")
if max_attempts is not None:
check_value(max_attempts, (1, FLOAT_MAX_INTEGER))
def new_method(self, *args, **kwargs):
[transforms, num_ops], _ = parse_user_args(method, *args, **kwargs)
type_check(num_ops, (int,), "num_ops")
check_positive(num_ops, "num_ops")
if num_ops > len(transforms):
raise ValueError("num_ops is greater than transforms list size.")
parsed_transforms = []
for op in transforms:
if op and getattr(op, 'parse', None):
parsed_transforms.append(op.parse())
else:
parsed_transforms.append(op)
type_check(parsed_transforms, (list, tuple,), "transforms")
for index, arg in enumerate(parsed_transforms):
if not isinstance(arg, (TensorOp, TensorOperation)):
raise TypeError("Type of Transforms[{0}] must be c_transform, but got {1}".format(index, type(arg)))
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[degrees, translate, scale, shear, resample,
fill_value], _ = parse_user_args(method, *args, **kwargs)
check_degrees(degrees)
if translate is not None:
type_check(translate, (list, tuple), "translate")
type_check_list(translate, (int, float), "translate")
if len(translate) != 2 and len(translate) != 4:
raise TypeError(
"translate should be a list or tuple of length 2 or 4.")
for i, t in enumerate(translate):
check_value(t, [-1.0, 1.0], "translate at {0}".format(i))
if scale is not None:
type_check(scale, (tuple, list), "scale")
type_check_list(scale, (int, float), "scale")
if len(scale) == 2:
if scale[0] > scale[1]:
raise ValueError(
"Input scale[1] must be equal to or greater than scale[0]."
)
check_range(scale, [0, FLOAT_MAX_INTEGER])
check_positive(scale[1], "scale[1]")
else:
raise TypeError("scale should be a list or tuple of length 2.")
if shear is not None:
type_check(shear, (numbers.Number, tuple, list), "shear")
if isinstance(shear, numbers.Number):
check_positive(shear, "shear")
else:
type_check_list(shear, (int, float), "shear")
if len(shear) not in (2, 4):
raise TypeError("shear must be of length 2 or 4.")
if len(shear) == 2 and shear[0] > shear[1]:
raise ValueError(
"Input shear[1] must be equal to or greater than shear[0]"
)
if len(shear) == 4 and (shear[0] > shear[1]
or shear[2] > shear[3]):
raise ValueError(
"Input shear[1] must be equal to or greater than shear[0] and "
"shear[3] must be equal to or greater than shear[2].")
type_check(resample, (Inter, ), "resample")
if fill_value is not None:
check_fill_value(fill_value)
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[batch_size, alpha, is_single], _ = parse_user_args(method, *args, **kwargs)
type_check(is_single, (bool,), "is_single")
type_check(batch_size, (int,), "batch_size")
type_check(alpha, (int, float), "alpha")
check_value(batch_size, (1, FLOAT_MAX_INTEGER))
check_positive(alpha, "alpha")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[image_batch_format, alpha, prob], _ = parse_user_args(method, *args, **kwargs)
type_check(image_batch_format, (ImageBatchFormat,), "image_batch_format")
type_check(alpha, (int, float), "alpha")
type_check(prob, (int, float), "prob")
check_pos_float32(alpha)
check_positive(alpha, "alpha")
check_value(prob, [0, 1], "prob")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[distortion_scale, prob,
interpolation], _ = parse_user_args(method, *args, **kwargs)
type_check(distortion_scale, (float, ), "distortion_scale")
type_check(prob, (float, ), "prob")
check_value(distortion_scale, [0., 1.], "distortion_scale")
check_value(prob, [0., 1.], "prob")
type_check(interpolation, (Inter, ), "interpolation")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[cutoff, ignore], _ = parse_user_args(method, *args, **kwargs)
type_check(cutoff, (int, float), "cutoff")
check_value(cutoff, [0, 100], "cutoff")
if ignore is not None:
type_check(ignore, (list, tuple, int), "ignore")
if isinstance(ignore, int):
check_value(ignore, [0, 255], "ignore")
if isinstance(ignore, (list, tuple)):
for item in ignore:
type_check(item, (int,), "item")
check_value(item, [0, 255], "ignore")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[is_hwc], _ = parse_user_args(method, *args, **kwargs)
type_check(is_hwc, (bool, ), "is_hwc")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[transform, ratio], _ = parse_user_args(method, *args, **kwargs)
type_check(ratio, (float, int), "ratio")
check_value(ratio, [0., 1.], "ratio")
type_check(transform, (TensorOp,), "transform")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[rescale, shift], _ = parse_user_args(method, *args, **kwargs)
check_pos_float32(rescale)
type_check(shift, (numbers.Number,), "shift")
return method(self, *args, **kwargs)
def new_method(self, *args, **kwargs):
[prob], _ = parse_user_args(method, *args, **kwargs)
type_check(prob, (float, int,), "prob")
check_value(prob, [0., 1.], "prob")
return method(self, *args, **kwargs)