liveBook · Manning

Data augmentation

As explained in Inside Deep Learning: Math, Algorithms, Models

The practice of creating additional training examples by applying label-preserving transformations to existing data.

Data augmentation creates additional training examples by applying randomized, label-preserving transformations to existing images. It increases input diversity when collecting more data is costly. section 3.6

Choose transformations that preserve the data’s meaning:

  • Use rotations, translations, shears, crops, perspective changes, or color adjustments when their labels remain valid. section 3.6
  • Randomize transform parameters; control optional transforms with probabilities such as p=0.5 or p=0.15. section 3.6
  • Inspect transformed samples before training: flipping MNIST digits can change a 9 into a 6. section 3.6

Training and testing pipelines

A PyTorch pipeline applies RandomAffine during training, then converts images with ToTensor; testing uses only ToTensor. A CNN comparison evaluates accuracy with and without augmentation.

How Inside Deep Learning: Math, Algorithms, Models covers it

How deep it goes
Explains randomized transforms, probabilities, label preservation, and PyTorch pipelines, but stops short of transformation theory.
Examples
Uses MNIST affine transforms, sample transformation visualizations, and accuracy curves comparing augmented and unaugmented CNNs.
What it assumes
Assumes image data and a PyTorch CNN workflow while teaching augmentation motivation, random transforms, probabilities, and label preservation.
What you can do after
Apply it via torchvision.transforms by composing training augmentations and evaluating the resulting CNN.
How it connects
Connects data augmentation with convolutional neural networks, pooling, image transformations, and model accuracy.
Cover of Inside Deep Learning: Math, Algorithms, Models
liveBookmanning.com
Read Inside Deep Learning: Math, Algorithms, Models on liveBook Get the book at manning.com

Also explained in

Compare how 9 books explain data augmentation →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Inside Deep Learning: Math, Algorithms, Models. The section links point to the parts of the book it draws on.