pub fn normalize(
input: &ArrayD<f64>,
mean: &[f64],
std: &[f64],
) -> Result<ArrayD<f64>, AugmentationError>Expand description
Normalize input: (x − mean[c]) / std[c].
For a [B, C, H, W] or [C, H, W] tensor the normalization is per-channel.
If mean and std have length 1 the same value is applied to all channels.
For 1-D or 2-D tensors element-wise normalization uses the first element of mean/std.