RotationPredictionLoss<T>

Models & Types · AiDotNet.LossFunctions

Self-supervised loss function based on rotation prediction for images.

Inheritance: LossFunctionBase<T> → RotationPredictionLoss<T>

Implements: ISelfSupervisedLoss<T>

For Beginners

This teaches the model to understand image structure without labels.

Imagine showing someone 100 photos, each rotated randomly:

  • They learn to recognize: which way is “up”, spatial relationships, object orientations
  • They don’t need to know: what the objects are (no labels needed)

After this training, when you show them 5 labeled cat photos:

  • They already understand image structure
  • They just need to learn: “cats look like THIS”
  • Much faster than learning everything from scratch!

How it works:

  1. Take each unlabeled image
  2. Create 4 versions: rotated by 0°, 90°, 180°, 270°
  3. Label each version: 0, 1, 2, 3 (which rotation was applied)
  4. Train model to predict the rotation

What the model learns:

  • Edge orientations
  • Spatial relationships
  • Object structure
  • “Natural” vs “unnatural” orientations

These features are very useful for actual classification tasks!

How It Works

Rotation prediction is a self-supervised task where:

  1. Images are rotated by 0°, 90°, 180°, or 270°
  2. Model predicts which rotation was applied (4-class classification)
  3. Model learns spatial relationships and features without needing class labels

Example

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

var loss = new RotationPredictionLoss<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($"RotationPredictionLoss = {value:F4}");

Constructors

Constructor Summary
RotationPredictionLoss()

Methods

Method Summary
T CalculateLoss(Vector<T> predicted, Vector<T> target) Categorical cross-entropy loss for rotation prediction (4-class).
Tensor<T> ComputeTapeLoss(Tensor<T> predicted, Tensor<T> target)
ValueTuple<TInput, TOutput> CreateTask(TInput input)