What is test-time augmentation (TTA)?

Test-time augmentation (TTA) is a technique used in computer vision and machine learning, where during inference (testing), various augmentations (such as rotations, reflections, scaling) are applied to the input data. The model makes predictions for each transformed version, and then the results are combined (for example, averaged) to obtain a more robust and accurate output.

TTA helps improve prediction quality by accounting for data variations that the model might not have fully captured during training.

Example: multiple rotations and reflections are applied to an image, predictions are obtained for each, and then class probabilities are averaged.