mindspore.ops.MaxUnpool3D

class mindspore.ops.MaxUnpool3D(ksize, strides=0, pads=0, output_shape=(), data_format='NCDHW')

Computes the inverse of mindspore.ops.MaxPool3D.

MaxUnpool3D keeps the maximal value and set all position of non-maximal values to zero. Typically the input is of shape $(N,C,D_{in},H_{in},W_{in})$, the output is of shape $(N,C,D_{out},H_{out},W_{out})$, the operation is as follows.

$$\begin{array}{r}\begin{array}{l}\\ D_{out}=(Din-1)\times strides[0]-2\times pads[0]+ksize[0]\\ H_{out}=(Hin-1)\times strides[1]-2\times pads[1]+ksize[1]\\ W_{out}=(Win-1)\times strides[2]-2\times pads[2]+ksize[2]\end{array}\end{array}$$

Warning

This is an experimental API that is subject to change or deletion.

Parameters

• ksize (Union[int, tuple[int]]) – The size of kernel used to take the maximum value, is an int number that represents depth, height and width of the kernel, or a tuple of three int numbers that represent depth, height and width respectively.

• strides (Union[int, tuple[int]], optional) –

  The distance of kernel moving. Default: 0 .

  • If it is an int number, the depth, height and width of movement are all equal to strides.

  • If it is a tuple of three int numbers, they represent depth, height and width of movement respectively.

  • If strides is 0 or (0, 0, 0), then strides equal to ksize.

• pads (Union[int, tuple[int]], optional) –

  The pad value to be filled. Default: 0 .

  • If pads is an integer, the paddings of depth, height and width are the same, equal to pads.

  • If pads is a tuple of three integers, the padding of depth, height and width equal to pads[0], pads[1] and pads[2] correspondingly.

• output_shape (tuple[int], optional) – The target output size. Default: () . If $output\mathrm{\_}shape==()$, then the shape of output computed by kszie, strides and pads shown above. If $output\mathrm{\_}shape!=()$, then output_shape format must be $(N,C,D,H,W)$ or $(N,D,H,W,C)$ and output_shape must be in range $[(N,C,D_{out}-strides[0],H_{out}-strides[1],W_{out}-strides[2]),(N,C,D_{out}+strides[0],H_{out}+strides[1],W_{out}+strides[2])]$.

• data_format (str, optional) – The optional value for data format. Currently support 'NCDHW' and 'NDHWC' . Default: 'NCDHW' .

Inputs:

• x (Tensor) - The input Tensor to invert. Tensor of shape $(N,C,D_{in},H_{in},W_{in})$ or $(N,D_{in},H_{in},W_{in},C)$.

• argmax (Tensor) - Max values' index. Tensor that has the same shape as x. Values of argmax must be in range $[0,D_{in}\times H_{in}\times W_{in}-1]$. Data type must be int32 or int64.

Outputs:

Tensor, with shape $(N,C,D_{out},H_{out},W_{out})$ or $(N,D_{out},H_{out},W_{out},C)$. Has the same data type with x.

Raises

• TypeError – If data type of x or argmax is Number.

• TypeError – If ksize, strides or pads is neither int nor tuple.

• ValueError – If numbers in strides or ksize is negative.

• ValueError – If numbers in pads is negative.

• ValueError – If ksize, strides or pads is a tuple whose length is not equal to 3.

• ValueError – If data_format is not a str or is neither 'NCDHW' nor 'NDHWC'.

• ValueError – If output_shape whose length is neither 0 or 5.

• ValueError – If output_shape is not close to output size range computed by attr ksize, strides, pads.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import numpy as np
>>> from mindspore import Tensor, ops
>>> x = Tensor(np.array([[[[[0, 1], [8, 9]]]]]).astype(np.float32))
>>> argmax = Tensor(np.array([[[[[0, 1], [2, 3]]]]]).astype(np.int64))
>>> maxunpool3d = ops.MaxUnpool3D(ksize=1, strides=1, pads=0)
>>> output = maxunpool3d(x, argmax)
>>> print(output.asnumpy())
[[[[[0. 1.]
    [8. 9.]]]]]