MaxUnpool

Description

Computes the partial inverse operation of the MaxPool operation. The input of this operation is usually the output of the MaxPool operation. The first input Tensor x is the tensor to be unpooled. It is usually the first output (pooling tensor) of MaxPool. The second input tensor i contains the index of the (locally largest) element corresponding to the element in the first input tensor x. The input tensor i is usually the second output of the MaxPool operation. The third (optional) input is a tensor that specifies the output size of the unpooling operation.

Input

• x: input tensor of type float16 or float.
• i: input tensor of type int64.
• output_shape: (optional) output shape. The data type is int64.

Attribute

• kernel_shape (mandatory): a list of int or int values, specifying the kernel size along each axis. The data type is INTS.
• pads: a list of int or int values, specifying the padding size along each axis. The data type is INTS.
• strides: a list of int or int values, specifying the stride size along each axis. The data type is INTS.

Output

y: output tensor of the identical data type as input x.

Constraints

None.

ONNX Opset Support

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