InstanceNormalization

Description

Compute y = scale * (x - mean) / sqrt(variance + epsilon) + B, where mean and variance are the mean and method of each channel per instance.

Input

Three inputs:

x: tensor of type float16 or float.

scale: 1D tensor. Its dimension is the same as the length of the C axis of x. It has the same data type as input x.

B: 1D tensor. Its dimension is the same as the length of the C axis of x. Its data type is the same as that of input x.

Output

One output:

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

Attribute

epsilon: float. Defaults to 1e-05 to avoid division by zero.

ONNX Opset Support

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