Struct StyleTransferLoss 

pub struct StyleTransferLoss {
    pub weights: StyleTransferWeights,
}

Struct that combines content, style, and TV losses with configurable weights.

Example

use scirs2_vision::style_transfer::{StyleTransferLoss, StyleTransferWeights, gram_matrix};
use scirs2_core::ndarray::Array3;

let img: Array3<f64> = Array3::ones((3, 8, 8));
let gen_gram = gram_matrix(&img);
let sty_gram = gram_matrix(&img);

let loss = StyleTransferLoss::new(StyleTransferWeights::default());
let v = loss.total(&gen_gram, &sty_gram, &img, &img);
assert!(v >= 0.0);

Fields

weights: StyleTransferWeights

Weights for each loss component.

Implementations

impl StyleTransferLoss

pub fn new(weights: StyleTransferWeights) -> Self

Create a new combined loss with the supplied weights.

pub fn total( &self, generated_gram: &Array2<f64>, style_gram: &Array2<f64>, generated: &Array3<f64>, content: &Array3<f64>, ) -> f64

Compute the total weighted loss.

Arguments

• generated_gram – Gram matrix of the generated image’s features.
• style_gram – Gram matrix of the style image’s features.
• generated – Generated feature map (for content + TV).
• content – Content feature map.

pub fn components( &self, generated_gram: &Array2<f64>, style_gram: &Array2<f64>, generated: &Array3<f64>, content: &Array3<f64>, ) -> (f64, f64, f64)

Compute individual loss components without combining.

Returns (content_loss, style_loss, tv_loss).

Trait Implementations

impl Clone for StyleTransferLoss

fn clone(&self) -> StyleTransferLoss

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for StyleTransferLoss

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

Formats the value using the given formatter.