I wrote a long practical guide on image augmentation based on ~10 years of train...

I wrote a long practical guide on image augmentation based on ~10 years of training computer vision models and ~7 years maintaining Albumentations.

Despite augmentation being used everywhere, most discussions are still very surface-level (“flip, rotate, color jitter”).

In this article I tried to go deeper and explain:

• The *two regimes of augmentation*: – in-distribution augmentation (simulate real variation) – out-of-distribution augmentation (regularization)

• Why *unrealistic augmentations can actually improve generalization*

• How augmentation relates to the *manifold hypothesis*

• When and why *Test-Time Augmentation (TTA)* helps

• Common *failure modes* (label corruption, over-augmentation)

• How to design a *baseline augmentation policy that actually works*

The guide is long but very practical — it includes concrete pipelines, examples, and debugging strategies.

Would love feedback from people working on real CV systems.