plenoptic.load_images

plenoptic.load_images(paths, as_gray=True, sorted_key=None)

Load in images.

Our models and synthesis methods generally expect their inputs to be 4d float32 images: (batch, channel, height, width), where the batch dimension contains multiple images and channel contains something like RGB or color channel. This function helps you get your inputs into that format. It accepts either a single file, a list of files, or a single directory containing images, will load them in, normalize them to lie between 0 and 1, convert them to float32, optionally convert them to grayscale, make them tensors, and get them into the right shape.

Parameters:

• paths (str | list[str] | Path | list[Path]) – A str or list of strs. If a list, must contain paths of image files. If a str, can either be the path of a single image file or of a single directory. If a directory, we try to load every file it contains (using imageio.imread) and skip those we cannot (thus, for efficiency you should not point this to a directory with lots of non-image files). This is NOT recursive.

• as_gray (bool (default: True)) – Whether to convert the images into grayscale or not after loading them. If False, we do nothing. If True, we call skimage.color.rgb2gray on them, which will result in a single channel.

• sorted_key (None | Callable (default: None)) – How to sort the images. If None and paths is a directory, will sort the paths alphabetically. If paths is a list of files, must be None and is ignored. See Sorting Techniques for details on other possible values, and note that the objects to sort are pathlib.Path objects.

Return type:

Tensor

Returns:

images – 4d tensor containing the images.

Raises:

• FileNotFoundError – If any of the explicit image paths do not exist.

• ValueError – If the images we attempt to load are not all the same shape.

• ValueError – If paths is a single file or list of files and sorted_key is not None.

Warns:

UserWarning – If paths is a directory and any of the files it contains are non-images.

Examples

When sorted_key=None, images from a directory are sorted alphabetically by filename.

>>> import plenoptic as po
>>> img_dir = po.data.fetch_data("test_images.tar.gz") / "256"
>>> titles = ["color_wheel", "curie", "einstein", "metal", "nuts"]
>>> imgs = po.load_images(img_dir)
>>> po.plot.imshow(imgs, title=titles)
<PyrFigure size ... with 5 Axes>

(png, hires.png, pdf)

Sort the images by the second letter of their filename:

>>> import plenoptic as po
>>> img_dir = po.data.fetch_data("test_images.tar.gz") / "256"
>>> titles = ["metal", "einstein", "color_wheel", "curie", "nuts"]
>>> imgs = po.load_images(img_dir, sorted_key=lambda x: x.name[1])
>>> po.plot.imshow(imgs, title=titles)
<PyrFigure size ... with 5 Axes>

(png, hires.png, pdf)