mindspore.ops.space_to_batch_nd

mindspore.ops.space_to_batch_nd(input_x, block_size, paddings)

Divides a tensor's spatial dimensions into blocks and combines the block sizes with the original batch.

\[\begin{split}\begin{array}{ll} \\ n' = n*(block\_size[0] * ... * block\_size[M]) \\ w'_i = (w_i + paddings[i][0] + paddings[i][1])//block\_size[i] \end{array}\end{split}\]

Note

• This operation divides the spatial dimensions [1, …, M] of the input into blocks of size block_size and interleaves them into the batch dimension (default: dimension 0). Before splitting, the spatial dimensions are padded with zeros according to paddings.

• If the input shape is \((n, c_1, ... c_k, w_1, ..., w_M)\), then the output shape will be \((n', c_1, ... c_k, w'_1, ..., w'_M)\).

• If block_size is a tuple or list, the length of block_size is M corresponding to the number of spatial dimensions. If block_size is an int, the block size of M dimensions are the same, equal to block_size. M must be 2 on Ascend.

Parameters

• input_x (Tensor) – The input tensor, must be a 4-D tensor on Ascend.

• block_size (Union[list(int), tuple(int), int]) – Specifies the block size for spatial dimension division.

• paddings (Union[tuple, list]) – The padding size for each spatial dimension.

Returns

Tensor

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore
>>> block_size = [2, 2]
>>> paddings = [[0, 0], [0, 0]]
>>> input_x = mindspore.tensor([[[[1, 2], [3, 4]]]], mindspore.float32)
>>> output = mindspore.ops.space_to_batch_nd(input_x, block_size, paddings)
>>> print(output)
[[[[1.]]]
 [[[2.]]]
 [[[3.]]]
 [[[4.]]]]