RealESRGANLoss<T>

Models & Types · AiDotNet.LossFunctions

Combined loss function for Real-ESRGAN super-resolution training.

Inheritance: LossFunctionBase<T> → RealESRGANLoss<T>

For Beginners

This loss function guides Real-ESRGAN training by balancing three goals:

  1. L1 Loss (pixel accuracy): Makes sure each pixel is close to the ground truth.

Like comparing photos pixel-by-pixel.

  1. Perceptual Loss (looks right): Uses a pre-trained network (VGG) to compare

high-level features. Ensures the output “looks right” even if pixels aren’t exact.

  1. GAN Loss (realistic details): The discriminator judges if output looks real.

This adds fine details and textures that make images look natural.

The weights control how much each goal matters:

  • Higher L1 weight = more pixel-accurate but potentially blurry
  • Higher perceptual weight = better visual quality
  • Higher GAN weight = more realistic textures but potential artifacts

The default weights (1.0, 1.0, 0.1) are from the Real-ESRGAN paper.

How It Works

Real-ESRGAN uses a combination of three loss functions for training:

  • L1 (pixel-wise) loss: Ensures pixel-level accuracy
  • Perceptual (VGG) loss: Ensures perceptual quality using deep features
  • GAN (adversarial) loss: Ensures realistic details and textures

The total loss is computed as:

Reference: Wang et al., “Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data”, ICCV 2021. https://arxiv.org/abs/2107.10833

Example

using AiDotNet.LossFunctions;
using AiDotNet.Tensors.LinearAlgebra;

var loss = new RealESRGANLoss<float>();
var predicted = new Vector<float>(new[] { 0.9f, 0.2f, 0.7f });
var actual = new Vector<float>(new[] { 1.0f, 0.0f, 1.0f });

float value = loss.CalculateLoss(predicted, actual);
Console.WriteLine($"RealESRGANLoss = {value:F4}");

Constructors

Constructor Summary
RealESRGANLoss(double l1Weight, double perceptualWeight, double ganWeight, Func<Tensor<T>, Tensor<T>> featureExtractor) Initializes a new instance of the RealESRGANLoss class.

Properties

Property Summary
double GANWeight { get; } Gets the GAN loss weight.
double L1Weight { get; } Gets the L1 weight.
double PerceptualWeight { get; } Gets the perceptual loss weight.

Methods

Method Summary
T CalculateCombinedLoss(Vector<T> predicted, Vector<T> actual, Vector<T> discriminatorOutput) Calculates the full combined loss including GAN component.
T CalculateDiscriminatorLoss(Vector<T> realOutput, Vector<T> fakeOutput) Calculates the discriminator loss.
T CalculateGeneratorGANLoss(Vector<T> discriminatorOutput) Calculates the generator’s GAN loss component.
T CalculateLoss(Vector<T> predicted, Vector<T> actual) Calculates the combined Real-ESRGAN loss.
Tensor<T> ComputeTapeLoss(Tensor<T> predicted, Tensor<T> target)
void SetNetworks(INeuralNetwork<T> featureExtractor, INeuralNetwork<T> discriminator) Sets the feature extractor and discriminator networks for tape-based training.