liveBook · Manning

Image_dataset_from_directory

As explained in Deep Learning with R, Third Edition

image_dataset_from_directory is a Keras utility that constructs an image-input dataset by reading class-organized image files from a directory hierarchy.

image_dataset_from_directory constructs an image dataset from files arranged in directories. It finds class subdirectories, indexes their images, decodes JPEG files, resizes them, shuffles them, and returns batches suitable for a Keras model. section 8.2.4

Create training, validation, and test datasets

image_size <- shape(180, 180) batch_size <- 32 train_dataset <- image_dataset_from_directory(new_base_dir / "train", image_size = image_size, batch_size = batch_size) validation_dataset <- image_dataset_from_directory(new_base_dir / "validation", image_size = image_size, batch_size = batch_size) test_dataset <- image_dataset_from_directory(new_base_dir / "test", image_size = image_size, batch_size = batch_size) Each directory contains class subdirectories. The resulting datasets can be passed directly to fit(). section 8.2.4

The utility automates the main preparation steps required before image data reaches a model:

  • Reads image files from class-organized subdirectories and infers the classes from that organization. section 8.2.4
  • Decodes JPEG content into RGB pixel grids and converts the images into tensors. section 8.2.4
  • Resizes images to a shared size and packs them into batches, such as batches of 32 images. section 8.2.4
  • Supports unlabeled image collections with labels = NULL and can preserve aspect ratio through crop_to_aspect_ratio = TRUE. section 17.2.1
Listing 8.9 Using image_dataset_from_directory() to read images from directories
image_size <- shape(180, 180)
batch_size <- 32

train_dataset <-
  image_dataset_from_directory(new_base_dir / "train",
                               image_size = image_size,
                               batch_size = batch_size)
validation_dataset <-
  image_dataset_from_directory(new_base_dir / "validation",
                               image_size = image_size,
                               batch_size = batch_size)
test_dataset <-
  image_dataset_from_directory(new_base_dir / "test",
                               image_size = image_size,
                               batch_size = batch_size)

Use the complete listing to create separate datasets for training, validation, and testing. section 8.2.4

The returned object is a tf.data.Dataset: an iterable sequence of input batches and labels that can be inspected, transformed, and supplied directly to model training. section 8.2.4

How Deep Learning with R, Third Edition covers it

How deep it goes
Explains directory indexing, decoding, resizing, shuffling, batching, and dataset output without detailing implementation internals.
Examples
Uses class-organized train, validation, and test folders plus the unlabeled Oxford Flowers image collection.
What it assumes
Assumes image files exist on disk and teaches directory organization, decoding, resizing, tensor conversion, and batching.
What you can do after
Apply it via Keras to produce batched image tensors and labels ready for model fitting.
How it connects
Connects image preprocessing with Keras models, tf.data.Dataset, batching, validation, testing, and image generation.
Cover of Deep Learning with R, Third Edition
liveBookmanning.com
Read Deep Learning with R, Third Edition on liveBook Get the book at manning.com

Also explained in

Compare how 2 books explain image_dataset_from_directory →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with R, Third Edition. The section links point to the parts of the book it draws on.