Class ImageDataset

java.lang.Object

smile.vision.ImageDataset

All Implemented Interfaces:

AutoCloseable, Iterable<SampleBatch>, Dataset

public class ImageDataset
extends Object
implements Dataset

Each of these directories should contain one subdirectory for each class in the dataset. The subdirectories are named after the corresponding class and contain all the images for that class. Ensure that each image file is named uniquely and stored in a common format such as JPEG or PNG.

Constructor Summary

Constructors

Constructor	Description
ImageDataset(String root, int batch, Transform transform, ToIntFunction<String> targetTransform)	Constructor.
ImageDataset(ArrayList<Path> images, int batch, Transform transform, ToIntFunction<String> targetTransform)	Constructor.

Method Summary

Modifier and Type	Method	Description
void	close()
Iterator<SampleBatch>	iterator()
ImageDataset	sample(int size)	Returns a random sample of the dataset with the given size.
long	size()	Returns the size of dataset.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface Iterable

forEach, spliterator

Constructor Details

ImageDataset

public ImageDataset(String root,
 int batch,
 Transform transform,
 ToIntFunction<String> targetTransform)
             throws IOException

Constructor.

Parameters:

root - the root directory of image dataset.

batch - the mini-batch size.

transform - the transformation from image to tensor.

targetTransform - the transform from image label to class index.

Throws:

IOException - if the root directory doesn't exist or doesn't have images.

ImageDataset

public ImageDataset(ArrayList<Path> images,
 int batch,
 Transform transform,
 ToIntFunction<String> targetTransform)

Constructor.

Parameters:

images - the list of image files. The folder name of an image file will be used as the class label.

batch - the mini-batch size.

transform - the transformation from image to tensor.

targetTransform - the transform from image label to class index.

Method Details

sample

public ImageDataset sample(int size)

Returns a random sample of the dataset with the given size. The sample shares the same image files and transforms with the original dataset. The sample is useful for quick testing or debugging on a smaller subset of the data.

Parameters:

size - the sample size.

Returns:

a random sample of the dataset with the given size.

close

public void close()

Specified by:

close in interface AutoCloseable

size

public long size()

Description copied from interface: Dataset

Returns the size of dataset.

Specified by:

size in interface Dataset

Returns:

the size of dataset.

iterator

public Iterator<SampleBatch> iterator()

Specified by:

iterator in interface Iterable<SampleBatch>