Data Augmentation
Domain: Data & Datasets
🎯 Motivation
Creating more training data by modifying existing data: flipping, rotating, paraphrasing. Increases dataset size and diversity.
📋 Concrete Example
Augmentation: 1 photo of a cat -> flip, rotate, crop, adjust brightness = 10 photos. Dataset grows from 1K to 10K images. More diverse training data.
🔗 Analogy
A chef creating variations of a dish - same ingredients, different presentations. Data augmentation creates variations of data for more robust training.