SpaceToDepth

Description

Rearranges the data in the spatial dimension of the input tensor to the depth dimension. That is, the data in the height and width dimensions is moved to the depth dimension.

Input

input: tensor of type float16 or float, with shape [N, C, H, W], where N indicates batch, C indicates channel or depth, H indicates height, and W indicates width.

Attribute

blocksize: int type, specifying the amount of data to be moved in the height and depth dimensions.

Output

output: tensor of the same data type as input, with shape [N, C * blocksize * blocksize, H / blocksize, W / blocksize].

Constraints

None.

ONNX Opset Support

Opset v8/v9/v10/v11/v12/v13/v14/v15/v16/v17/v18