Changelog

1.10

• Merged PR 18

  • Add support to CuPy as input

  • Add optional support for RAPIDS cuCIM

• Merged PR 25

  • Implement Integration Test checking for stability of GLCM output

• Merged PR 27

  • Implemented GLCM Feature Selection to optimize out unnecessary GLCM Stages

• Merged PR 29

  • Fix issue with CuPy ndarray incompatible with tqdm

• Merged PR 30

  • Add Batch Processing for vanilla GLCM

    • GLCMCross doesn’t perform faster with this.

• Merged PR 32

  • Author: Julio Faracco

  • Implement Dissimilarity Feature

• Merged PR 35

  • Adjusts PR 18 coerces inputs to CuPy if NumPy.

  • Reduces duplicate code for handling NumPy and CuPy arrays conditionally.

• Merged PR 36

  • Implement NaN Handling by ignoring contribution to GLCM

• Merged PR 41

  • Resolved issue where NA replacement value was incorrectly set to bin_from

  • Resolved issue where NA summing used .sum instead of np.nansum

• Merged PR 42

  • Optimize kernel, allow thread early exit if Probability p==0

  • Fix List / Set like typehints not flexible enough. Migrate to typing.Sequence

1.9

• Merged PR 10

  • Resolved Issue #9

  • Add bad ndim raise

  • Improve raise message on bad ndim

• Merged PR 14

  • Resolved Issue #11

  • Allow tqdm to be silenced

• Merged PR 15

  • Resolved Issue #13

  • normalize_features -> normalized_features

  • Fix glcm and glcm_cross unexpected arg order behaviour

  • Remove test__from_windows() as it’s redundant

  • Fix test_from_2d_image failing due to missing 3rd dimension

  • Remove test_image_tiff as it’s redundant

  • Fix reference GLCM._binner to binner in utils

1.8

• Merged PR 7

  • Implement Cross GLCM

1.7

• Fix issue with GLCM overflowing by making it float32

1.6

• Dropped dependency on J variables as I & J are always the same