Image Augmentation

Augmentation is defined as making something better by making it greater in size or amount. This is exactly what data or image augmentation does. You use augmentation to provide the model with more versions of your image training data. Remember that the more data you have, the better the model's performance will be. By augmenting your data, you can transform your images in a way that makes the model generalize better on real data. To do this, you transform the images that you have at your disposal so that you can use your augmented images alongside your original image dataset to train with a greater variation and variety than you would have otherwise. This improves results and prevents overfitting. Take a look at the following three images:

Figure 7.14: Augmented leopard images

It's clear that this is the same leopard in all three images. They're just in different positions. Neural networks can still make sense of this due to...