module onnxrt.ops_cpu.op_resize
#
Short summary#
module mlprodict.onnxrt.ops_cpu.op_resize
Runtime operator.
Classes#
class |
truncated documentation |
---|---|
Resize ====== Resize the input tensor. In general, it calculates every value in the output tensor as a weighted average … |
Functions#
function |
truncated documentation |
---|---|
From https://stackoverflow.com/a/1235363 Generate a cartesian product of input arrays. Parameters ———- … |
|
Properties#
property |
truncated documentation |
---|---|
|
Returns the list of arguments as well as the list of parameters with the default values (close to the signature). … |
|
Returns the list of modified parameters. |
|
Returns the list of optional arguments. |
|
Returns the list of optional arguments. |
|
Returns all parameters in a dictionary. |
Methods#
method |
truncated documentation |
---|---|
Documentation#
Runtime operator.
- class mlprodict.onnxrt.ops_cpu.op_resize.Resize(onnx_node, desc=None, expected_attributes=None, **options)#
Bases:
OpRun
Resize the input tensor. In general, it calculates every value in the output tensor as a weighted average of neighborhood (a.k.a. sampling locations) in the input tensor. Each dimension value of the output tensor is: <br/>
output_dimension = floor(input_dimension * (roi_end - roi_start) * scale) <br/>
if input "sizes" is not specified.
Attributes
antialias: If set to 1, “linear” and “cubic” interpolation modes will use an antialiasing filter when downscaling. Antialiasing is achieved by stretching the resampling filter by a factor max(1, 1 / scale), which means that when downsampling, more input pixels contribute to an output pixel. Default value is
nameantialiasi0typeINT
(INT)axes: If provided, it specifies a subset of axes that ‘roi’, ‘scales’ and ‘sizes’ refer to. If not provided, all axes are assumed [0, 1, …, r-1], where r = rank(data). Non-specified dimensions are interpreted as non-resizable. Negative value means counting dimensions from the back. Accepted range is [-r, r-1], where r = rank(data). Behavior is undefined if an axis is repeated. default value cannot be automatically retrieved (INTS)
coordinate_transformation_mode:
This attribute describes how to transform the coordinate in the resized tensor to the coordinate in the original tensor. <br/>
The coordinate of each dimension is transformed individually. Let’s describe a case using axis x as an example. Denote x_resized as the coordinate of axis x in the resized tensor, x_original as the coordinate of axis x in the original tensor, length_original as the length of the original tensor in axis x, length_resized as the length of the resized tensor in axis x, roi_x = (start_x, end_x) of the axis x in input “roi”, scale = length_resized / length_original, <br/>
if coordinate_transformation_mode is “half_pixel”, <br/> x_original = (x_resized + 0.5) / scale - 0.5 <br/>
if coordinate_transformation_mode is “pytorch_half_pixel”, <br/> x_original = length_resized > 1 ? (x_resized + 0.5) / scale - 0.5 : 0 <br/>
if coordinate_transformation_mode is “align_corners”, <br/> x_original = x_resized * (length_original - 1) / (length_resized - 1) <br/>
if coordinate_transformation_mode is “asymmetric”, <br/> x_original = x_resized / scale <br/>
if coordinate_transformation_mode is “tf_crop_and_resize”, <br/> x_original = length_resized > 1 ? start_x * (length_original - 1) + x_resized * (end_x - start_x) * (length_original - 1) / (length_resized - 1) : 0.5 * (start_x + end_x) * (length_original - 1) . Default value is
namecoordinatetransformationmodeshalfpixeltypeSTRING
(STRING) * cubic_coeff_a: The coefficient ‘a’ used in cubic interpolation. Two common choice are -0.5 (in some cases of TensorFlow) and -0.75 (in PyTorch). Check out Equation (4) in https://ieeexplore.ieee.org/document/1163711 for the details. This attribute is valid only if mode is “cubic”. Default value isnamecubiccoeffaf-0.75typeFLOAT
(FLOAT) * exclude_outside: If set to 1, the weight of sampling locations outside the tensor will be set to 0 and the weight will be renormalized so that their sum is 1.0. The default value is 0. Default value isnameexcludeoutsidei0typeINT
(INT) * extrapolation_value: When coordinate_transformation_mode is “tf_crop_and_resize” and x_original is outside the range [0, length_original - 1], this value is used as the corresponding output value. Default is 0.0f. Default value isnameextrapolationvaluef0.0typeFLOAT
(FLOAT) * keep_aspect_ratio_policy: This attribute describes how to interpret the sizes input with regard to keeping the original aspect ratio of the input, and it is not applicable when the scales input is used. <br/>Given a set of sizes, associated with a subset of axes (explicitly provided or default), and assuming d = axes[i], with i being the index of the provided sizes. <br/>
If keep_aspect_ratio_policy is “stretch”, the original aspect ratio is disregarded, and the input is resized to the specified size: <br/> out_size[d] = sizes[i] <br/>
If keep_aspect_ratio_policy is “not_larger”, the sizes are adjusted so that no extent of the output is larger than the specified size, while keeping the original aspect ratio: <br/> scale = Min(sizes[i] / in_size[d]) <br/> out_size[d] = round_int(scale * in_size[i]) <br/>
If keep_aspect_ratio_policy is “not_smaller”, the sizes are adjusted so that no extent of the output is smaller than the specified size, while keeping the original aspect ratio: <br/> scale = Max(sizes[i] / in_size[d]) <br/> out_size[d] = round_int(scale * in_size[i]) <br/>
For non-resizable axes (those not specified in axes), the output size will be equal to the input size.
Note: round_int stands for computing the nearest integer value, rounding halfway cases up. Default value is
namekeepaspectratiopolicysstretchtypeSTRING
(STRING) * mode: Three interpolation modes: “nearest” (default), “linear” and “cubic”. The “linear” mode includes linear interpolation for 1D tensor and N-linear interpolation for N-D tensor (for example, bilinear interpolation for 2D tensor). The “cubic” mode includes cubic interpolation for 1D tensor and N-cubic interpolation for N-D tensor (for example, bicubic interpolation for 2D tensor). Default value isnamemodesnearesttypeSTRING
(STRING) * nearest_mode: Four modes: “round_prefer_floor” (default, as known as round half down), “round_prefer_ceil” (as known as round half up), “floor”, “ceil”. Only used by nearest interpolation. It indicates how to get “nearest” pixel in input tensor from x_original, so this attribute is valid only if “mode” is “nearest”. Default value isnamenearestmodesroundpreferfloortypeSTRING
(STRING)Inputs
Between 1 and 4 inputs.
X (heterogeneous)T1: N-D tensor
roi (optional, heterogeneous)T2: 1-D tensor given as [start1, …, startN, end1, …, endN], where N is the rank of X or the length of axes, if provided. The RoIs’ coordinates are normalized in the coordinate system of the input image. It only takes effect when coordinate_transformation_mode is “tf_crop_and_resize”
scales (optional, heterogeneous)tensor(float): The scale array along each dimension. It takes value greater than 0. If it’s less than 1, it’s sampling down, otherwise, it’s upsampling. The number of elements of ‘scales’ should be the same as the rank of input ‘X’ or the length of ‘axes’, if provided. One of ‘scales’ and ‘sizes’ MUST be specified and it is an error if both are specified. If ‘sizes’ is needed, the user can use an empty string as the name of ‘scales’ in this operator’s input list.
sizes (optional, heterogeneous)tensor(int64): Target size of the output tensor. Its interpretation depends on the ‘keep_aspect_ratio_policy’ value.The number of elements of ‘sizes’ should be the same as the rank of input ‘X’, or the length of ‘axes’, if provided. Only one of ‘scales’ and ‘sizes’ can be specified.
Outputs
Y (heterogeneous)T1: N-D tensor after resizing
Type Constraints
T1 tensor(uint8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(int8), tensor(int16), tensor(int32), tensor(int64), tensor(bfloat16), tensor(float16), tensor(float), tensor(double), tensor(string), tensor(bool), tensor(complex64), tensor(complex128): Constrain input ‘X’ and output ‘Y’ to all tensor types.
T2 tensor(float16), tensor(float), tensor(double): Constrain roi type to float or double.
Version
Onnx name: Resize
This version of the operator has been available since version 18.
Runtime implementation:
Resize
- __init__(onnx_node, desc=None, expected_attributes=None, **options)#
- _run(X, roi, scales=None, sizes=None, attributes=None, verbose=0, fLOG=None)#
Should be overwritten.
- mlprodict.onnxrt.ops_cpu.op_resize._cartesian(arrays, out=None)#
From https://stackoverflow.com/a/1235363 Generate a cartesian product of input arrays. Parameters ———- arrays : list of array-like
1-D arrays to form the cartesian product of.
- outndarray
Array to place the cartesian product in.
Returns#
- outndarray
2-D array of shape (M, len(arrays)) containing cartesian products formed of input arrays.
Examples#
>>> cartesian(([1, 2, 3], [4, 5], [6, 7])) array([[1, 4, 6], [1, 4, 7], [1, 5, 6], [1, 5, 7], [2, 4, 6], [2, 4, 7], [2, 5, 6], [2, 5, 7], [3, 4, 6], [3, 4, 7], [3, 5, 6], [3, 5, 7]])
- mlprodict.onnxrt.ops_cpu.op_resize._cubic_coeffs(ratio, A=-0.75)#
- mlprodict.onnxrt.ops_cpu.op_resize._get_all_coords(data)#
- mlprodict.onnxrt.ops_cpu.op_resize._get_neighbor(x, n, data)#
- mlprodict.onnxrt.ops_cpu.op_resize._get_neighbor_idxes(x, n, limit)#
- mlprodict.onnxrt.ops_cpu.op_resize._interpolate_1d_with_x(data, scale_factor, x, get_coeffs, roi=None, extrapolation_value=0.0, coordinate_transformation_mode=b'half_pixel', exclude_outside=False)#
- mlprodict.onnxrt.ops_cpu.op_resize._interpolate_nd(data, get_coeffs, output_size=None, scale_factors=None, roi=None, **kwargs)#
- mlprodict.onnxrt.ops_cpu.op_resize._interpolate_nd_with_x(data, n, scale_factors, x, get_coeffs, roi=None, **kwargs)#
- mlprodict.onnxrt.ops_cpu.op_resize._linear_coeffs(ratio)#
- mlprodict.onnxrt.ops_cpu.op_resize._nearest_coeffs(ratio, mode=b'round_prefer_floor')#