Overview
Data augmentation helps prevent overfitting and improves model generalization. It is especially common in computer vision and audio processing.
Common Techniques
- Images: Rotating, flipping, cropping, or changing the brightness of photos.
- Text: Replacing words with synonyms or back-translating sentences.
- Audio: Adding background noise or changing the pitch.
Benefit
It effectively 'multiplies' your dataset, allowing you to train larger models with less original data.