Struct Normalization 

pub struct Normalization { /* private fields */ }

Per-channel (x - mean) / std on 0–255 pixel values, in RGB order.

It applies to the colour channels of the destination; an alpha channel always maps as NormalizationPreset::Unsigned. A Grey destination uses the first channel’s coefficients.

Implementations

impl Normalization

pub const fn preset(p: NormalizationPreset) -> Self

The coefficients of a named preset.

pub const fn unsigned() -> Self

x / 255.

pub const fn signed() -> Self

x / 127.5 - 1.

pub const fn raw() -> Self

x unchanged.

pub const fn imagenet() -> Self

ImageNet mean [123.675, 116.28, 103.53], std [58.395, 57.12, 57.375].

pub fn mean_std(mean: &[f32], std: &[f32]) -> Result<Self>

Explicit coefficients on the 0–255 scale: one value (every channel) or three (RGB).

Errors

Error::InvalidArgument for another count, a non-finite mean, or a std that is not finite and positive.

use edgefirst_tensor::Normalization;

assert_eq!(Normalization::mean_std(&[0.0], &[255.0]).unwrap(), Normalization::unsigned());
assert!(Normalization::mean_std(&[0.0], &[0.0]).is_err());
assert!(Normalization::mean_std(&[0.0], &[-1.0]).is_err());

pub fn unit_mean_std(mean: &[f32], std: &[f32], scale: f64) -> Result<Self>

Coefficients on the 0–1 scale, as Hugging Face and torchvision write them; scaled by scale (usually 255) onto the 0–255 scale in f64.

Errors

As for mean_std.

use edgefirst_tensor::Normalization;

let n = Normalization::unit_mean_std(&[0.5], &[0.5], 255.0).unwrap();
assert_eq!(n, Normalization::signed());

pub fn mean(&self) -> [f32; 3]

The per-channel mean, RGB order, on the 0–255 scale.

pub fn std(&self) -> [f32; 3]

The per-channel standard deviation, RGB order, on the 0–255 scale.

pub fn as_preset(&self) -> Option<NormalizationPreset>

The preset with exactly these coefficients, if any.

Trait Implementations

impl Clone for Normalization

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for Normalization

impl Debug for Normalization

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Normalization

fn default() -> Self

Returns the “default value” for a type.

impl From<NormalizationPreset> for Normalization

fn from(p: NormalizationPreset) -> Self

Converts to this type from the input type.

impl PartialEq for Normalization

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for Normalization