Skip to content
BENIBIX Learn

Data Augmentation

Also known as Augmentation (Machine Learning), Image Augmentation

Make more examples by changing the ones you have. Flip them, crop them, tint them.

Think of it like

It is like a coach who throws the ball from new spots. The player learns to catch it from anywhere.

Example

Lucy has 500 photos of ripe apples. She flips, turns and dims each one four ways. Her model now trains on 2,500 photos.

Why it matters

New data is slow and dear to get and tag. This stretches a small set of data and cuts overfitting.

Definition

Data augmentation makes more training examples by changing the ones you have. For a photo, you can flip it, turn it, crop it or make it darker. For text, you can swap in words that mean the same. The label stays the same, because a flipped cat is still a cat. The model sees more variety without new data being collected. It helps the model cope with real, messy inputs.
On the right, one photo becomes four changed copies. That is the trick.

Common mistake

Making a change that alters the answer. A flipped 6 becomes a 9, so the label is now wrong.

Key takeaways

  1. 1
    Change the copy. Flip, crop, turn or tint it.
  2. 2
    Keep the label. Only changes that keep the answer.
  3. 3
    More variety. The model copes with messy inputs.

Compare

Learn it properly