Struct neuro::data::ImageDataSet

pub struct ImageDataSet { /* fields omitted */ }

Structure representing a collection of images.

A builder class is provided for ease of creation: ImageDataSetBuilder.

Methods

impl ImageDataSet

pub fn from_dir(
    path: &Path,
    image_size: (u32, u32),
    one_hot_encode: bool,
    valid_frac: Option<f64>,
    image_ops: ImageOps
) -> Result<ImageDataSet, Error>

Constructs an ImageDataSet from a directory tree.

The images must be in folders named after the corresponding class in a train top-level directory. Optionally, if a test directory exists, its content will be used to create a test set. For instance:

pets/
  train/
    cats/
      img1.jpg
      img2.jpg
      ...
    dogs/
      img1.jpg
      img2.jpg
      ...
  test/
    cats/
      img1.jpg
      img2.jpg
      ...
    dogs/
      img1.jpg
      img2.jpg
      ...

The images are resized to the given size using nearest-neighbor interpolation. The aspect ratio is not conserved.

Arguments

• path - The path to the top level directory containing the images.
• image_size - The height and width of the images.
• one_hot_encode - Flag indicating whether the labels are one hot encoded.
• valid_frac - The fraction of the data used for validation.
• image_ops - The collection of operations applied on the images.

pub fn print_classes(&self)

Prints the classes in the dataset.

pub fn load_image(
    path: &Path,
    size: (u32, u32),
    image_ops: &ImageOps
) -> Result<(Vec<PrimitiveType>, u8), DataSetError>

Loads a single image from a path.

Arguments

• path - The path to the image.
• size - The height and width of the image.
• image_ops - The collection of operations to apply on the image.

Return value

Tuple containing a vector with the image and the number of channels.

pub fn load_image_vec(
    paths: &[&Path],
    image_size: (u32, u32),
    image_ops: &ImageOps
) -> Result<Tensor, Error>

Loads the images from the paths.

The images are scaled by the factor used to create the dataset.

Arguments

• paths - A slice of paths to the images. Each path must point to an individual image.

pub fn image_ops(&self) -> &ImageOps

Trait Implementations

impl DataSet for ImageDataSet

fn input_shape(&self) -> Dim4

Returns the dimension of the samples.

fn output_shape(&self) -> Dim4

Returns the dimension of the labels.

fn num_train_samples(&self) -> u64

Returns the number of samples in the training set.

fn num_valid_samples(&self) -> u64

Returns the number of samples in the validation set.

fn classes(&self) -> Option<Vec<String>>

Returns the classes in the data set.

fn x_train(&self) -> &Tensor

Returns a reference to the training samples.

fn y_train(&self) -> &Tensor

Returns a reference to the training labels.

fn x_valid(&self) -> Option<&Tensor>

Returns a reference to the validation samples.

fn y_valid(&self) -> Option<&Tensor>

Returns a reference to the validation labels.

fn x_test(&self) -> Option<&Tensor>

Returns a reference to the test samples.

fn y_test(&self) -> Option<&Tensor>

Returns a reference to the test labels.

fn x_train_stats(&self) -> &Option<(Scaling, Tensor, Tensor)>

Returns a reference to the type of scaling that has been applied to the input features and the values used for the scaling.

fn y_train_stats(&self) -> &Option<(Scaling, Tensor, Tensor)>

Returns a reference to the type of scaling that has been applied to the output labels and the values used for the scaling.

impl Display for ImageDataSet

fn fmt(&self, f: &mut Formatter) -> Result

Formats the value using the given formatter.