kaishi.image.util
¶
Image utilities and helper functions.
Module Contents¶
-
kaishi.image.util.
swap_channel_dimension
(tensor)¶ Swap between channels_first and channels_last.
If ‘tensor’ has 4 elements, it’s assumed to be the shape vector. Otherwise, it’s assumed that it’s the actual tensor. Returns the edited shape vector or tensor.
- Parameters
tensor (numpy.array) – shape vector or tensor to have channel dimensions swapped
- Returns
altered input with the channel dimensions swapped
- Return type
numpy.array
-
kaishi.image.util.
validate_image_header
(filename: str)¶ Validate that an image has a valid header.
Returns True if valid, False if invalid.
- Parameters
filename (str) – name of file to analyze
- Returns
flag indicating whether header is valid (by using imghdr.what())
- Return type
bool
-
kaishi.image.util.
get_batch_dimensions
(self, batch_size: int, channels_first: bool = True, image_type: str = 'small_image')¶ Get dimensions of the batch tensor. Note that the ‘batch_size’ argument can be the full data set.
- Parameters
self (
kaishi.image.dataset.ImageDataset
) – image dataset objectbatch_size (int) – batch size
channels_first (bool) – flag indicating whether channels are first or last dimension in each image
image_type (str) – one of “small_image”, “thumbnail”, or “patch”
- Returns
batch dimesions (4D tuple)
- Return type
tuple