mindspore.mint.nn.functional.max_unpool2d

mindspore.mint.nn.functional.max_unpool2d(input, indices, kernel_size, stride=None, padding=0, output_size=None)

Computes the inverse of max_pool2d.

max_unpool2d keeps the maximal value and sets all positions of non-maximal values to zero. Typically the input is of shape \((N, C, H_{in}, W_{in})\) or \((C, H_{in}, W_{in})\), and the output is of shape \((N, C, H_{out}, W_{out})\) or \((C, H_{out}, W_{out})\). The operation is as follows.

\[\begin{split}\begin{array}{ll} \\ H_{out} = (H_{in} - 1) \times stride[0] - 2 \times padding[0] + kernel\_size[0] \\ W_{out} = (W_{in} - 1) \times stride[1] - 2 \times padding[1] + kernel\_size[1] \\ \end{array}\end{split}\]

Warning

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

Parameters:

• input (Tensor) – The input Tensor to invert. Tensor of shape \((N, C, H_{in}, W_{in})\) or \((C, H_{in}, W_{in})\).

• indices (Tensor) – Max values' index represented by the indices. Tensor of shape must be same as input input. Values of indices must belong to \([0, H_{in} \times W_{in} - 1]\). Data type must be int32 or int64.

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

• stride (Union[int, tuple[int]], optional) – The distance of kernel moving, an int number that represents the height and width of movement are both stride, or a tuple of two int numbers that represent height and width of movement respectively. Default: None , which indicates the moving step is kernel_size .

• padding (Union[int, tuple[int]], optional) – The padding value to be filled. Default: 0 . If padding is an integer, the paddings of height and width are the same, equal to padding. If padding is a tuple of two integers, the padding of height and width equal to padding[0] and padding[1] correspondingly.

• output_size (tuple[int], optional) – The target output size. Default: None . If output_size == (), then the shape of output computed by kernel_size, stride and padding. If output_size != (), then output_size must be \((N, C, H, W)\) , \((C, H, W)\) or \((H, W)\) and output_size must belong to \([(N, C, H_{out} - stride[0], W_{out} - stride[1]), (N, C, H_{out} + stride[0], W_{out} + stride[1])]\).

Returns:

Tensor, with shape \((N, C, H_{out}, W_{out})\) or \((C, H_{out}, W_{out})\), with the same data type as input.

Raises:

• TypeError – If data type of input or indices is not supported.

• TypeError – If kernel_size, stride or padding is neither an int nor a tuple.

• ValueError – If the length of kernel_size, stride or padding is not 1 or 2.

• ValueError – If numbers in kernel_size or stride are not positive.

• ValueError – If numbers in padding are negative.

• ValueError – If the shapes of input and indices are different.

• ValueError – If the length of input is not 3 or 4.

• ValueError – If the type of output_size is not tuple.

• ValueError – If the length of output_size is not 0, 2, 3 or 4.

• ValueError – If output_size is not close to output size computed by attr kernel_size, stride, padding.

Supported Platforms:

Ascend

Examples

>>> import numpy as np
>>> from mindspore import Tensor, mint
>>> input = Tensor(np.array([[[[0, 1], [8, 9]]]]).astype(np.float32))
>>> indices = Tensor(np.array([[[[0, 1], [2, 3]]]]).astype(np.int64))
>>> output = mint.nn.functional.max_unpool2d(input, indices, 1, stride=1, padding=0)
>>> print(output.asnumpy())
[[[[0. 1.]
   [8. 9.]]]]