Gregory Lee
Gregory Lee
closes #363 by bumping the minimum required scikit-image attn: @jakirkham, @ajschmidt8
**Describe the bug** Changes made in #276 break use of some thresholding functions if the user does not have scikit-image >= 0.19. We should either bump our minimum requirement to...
# Ideas for further improvements to separable convolution Listing some thoughts for further improvements on top of #355. ### reduce redundant data fetching It would be nice to adapt the...
closes #193 This PR adds `cucim.skimage.feature.match_descriptors`. It is a very straightforward adaptation of the scikit-image code, only substituting numpy->cupy. The only differences of note are: - when new enough CuPy...
This PR adds benchmarks for a few of the recently introduced functions. It also uses more realistic nD label object generation for the segmentation benchmarks.
A histogram-based median for 2D 8 and 16-bit integer images was implemented in #317. There are some areas where this could be further improved in the future. ### improved boundary...
As mentioned in https://github.com/rapidsai/cucim/pull/252#issue-1187379516, we should come up with a more solid heuristic for the `choose_conv_method` helper in CuPy to help it more appropriately decide when to use a convolution...
**Describe the bug** In a [recent CI run](https://gpuci.gpuopenanalytics.com/job/rapidsai/job/gpuci/job/cucim/job/prb/job/cucim-gpu-test/CUDA=11.5,GPU_LABEL=driver-495,LINUX_VER=ubuntu20.04,PYTHON=3.9/308/console) there are various warnings related to test folders. For example: ``` 13:14:48 /opt/conda/envs/rapids/conda-bld/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_place/lib/python3.9/site-packages/setuptools/command/build_py.py:153: SetuptoolsDeprecationWarning: Installing 'cucim.skimage.filters.tests' as data is deprecated, please list...
`hessian_matrix_eigenvalues` has an efficient code path for 2D, but for nD the current implementation is very memory-hungry. In this case it would be beneficial to divide up the eigenvalue computations...
#340 proposed some minor improvements to gradient computation. However, I think there is still potential for a much stronger improvement. Specifically, by creating an custom `ElementwiseKernel`, we could write all...