Image Utilities

This module holds functionality related to loading and saving images.

stransfer.img_utils.concat_images(im1, im2, dim=2)[source]

Simple wrapper function that concatenates two images along a given dimension.

Parameters

  • im1 –

  • im2 –

  • dim – the dimension we want to concatenate across. By default it is the third dimension (width if images only have 3 dims)

Return type

Tensor

Returns

tensor representing concatenated image

stransfer.img_utils.image_loader(image_path)[source]

Loads an image from image_path, and transforms it into a torch.Tensor

Parameters

image_path (str) – path to the image that will be loaded

Return type

Tensor

Returns

tensor representing the image

stransfer.img_utils.image_loader_transform(image)[source]

Transform a PIL.Image instance into a torch.Tensor one.

Parameters

image (<module 'PIL.Image' from '/home/travis/virtualenv/python3.7.1/lib/python3.7/site-packages/PIL/Image.py'>) – image to transform

Return type

Tensor

Returns

torch.Tensor representing the image

stransfer.img_utils.imshow(image_tensor, ground_truth_image=None, denormalize=True, path='out.bmp')[source]

Utility function to save an input image tensor to disk.

Parameters

  • image_tensor (Tensor) – the tensor representing the image to be saved

  • ground_truth_image (Optional[Tensor]) – another tensor, representing another image. If provided, it will be concatenated to the image_tensor across the width dimension and this result will be saved to disk

  • denormalize – whether or not to denormalize (using IMAGENET mean and std) the tensor before saving it to disk.

  • path – the path where the image will be saved

Return type

None