beexplainable.preprocessing package

Submodules

beexplainable.preprocessing.dataset_transformers module

beexplainable.preprocessing.parsers module

Library for parsing image inputs before feeding them into the model.

beexplainable.preprocessing.parsers.parse_image(file_id: str, img_lookup, img_w: int, img_h: int, root: str = './')[source]

Read image file and resize it to (img_w, img_h, 3).

Parameters

  • file_id (str) – ID of the image to read.

  • img_lookup (tf.lookup.StaticHashTable) – Lookup table assigning file ID to file name.

  • img_w (int) – Image width after resizing.

  • img_h (int) – Image height after resizing.

  • root (str) – Root path where the image is stored. Defaults to current folder ./.

Returns

Decoded image.

Return type

Tensor

beexplainable.preprocessing.parsers.parse_image_and_mask(file_id: str, img_lookup, img_w: int, img_h: int, root_img: str = './', root_mask: str = './masks/')[source]

Read image file, resize it to (img_w, img_h, 3) and read object mask and resize it to (img_w, img_h, 1).

Parameters

  • file_id (str) – ID of the image to read.

  • img_lookup (tf.lookup.StaticHashTable) – Lookup table assigning file ID to file name.

  • img_w (int) – Image width after resizing.

  • img_h (int) – Image height after resizing.

  • root_img (str) – Root path where the image is stored. Defaults to current folder ./.

  • root_mask (str) – Root path where the mask is stored. Defaults to current folder ./masks/.

Returns

Decoded image and binary mask

Return type

Tuple[Tensor, Tensor]

beexplainable.preprocessing.parsers.parse_to_dataset(file_ids: List[str], labels: List[int], dataset_map)[source]

Retrieves file names from file_ids, decodes and stores them along with labels in a tf.Dataset.

Parameters

  • file_ids (list of strings) – File IDs to parse images from (will be keys of a dictionary mapping to the file names).

  • labels (list of integers) – Class labels corresponding to the images.

  • dataset_map – Function that maps a tuple of filenames and labels to a tuple of parsed images and labels.

Returns

Dataset of decoded images and labels.

Return type

tf.Dataset

Module contents