[ROADMAP] Kornia v0.2.1

This is the roadmap for kornia v0.2.1 that will focus mainly in morphology and epipolar geometry.

Feel free to volunteer yourself if you are interested in trying out some items(they do not have to be on the list). Please, take a read on the CONTRIBUTING notes.

• Epipolar Geometry Propose stuff potentially from opencv.calib3d and opencv.sfm

  • symmetrical_epipolar_distance https://github.com/kornia/kornia/pull/283 @ducha-aiki
  • sampson_epipolar_distance https://github.com/kornia/kornia/pull/283 @ducha-aiki
  • find_fundamental @edgarriba
  • find_homography @edgarriba
  • compute_correspond_epilines @edgarriba

• Morphology see discussion: https://github.com/kornia/kornia/issues/293#issuecomment-546725336

  • dilation https://github.com/kornia/kornia/pull/292
  • erosion
  • open
  • close
  • skeleton
  • morphological edge detection
  • histogram https://github.com/kornia/kornia/issues/413

• KMeans

  • @subhadarship https://github.com/subhadarship/kmeans_pytorch/issues/2

• Differentiable drawing. To be defined. Please, propose.

• Model Zoo *. To be discussed, use torch.hub

  • HardNet : https://github.com/DagnyT/hardnet

Improvements

Added features from albumentations, please propose whatever feature you consider is missing or might relevant for next versions. Most functions are already available, or can be implemented using kornia core operators. Need a common interface.

• kornia.augmentation.

  • RandomMotionBlur https://github.com/kornia/kornia/issues/494
  • Grayscale @dvd42
  • ElasticTransformation [link]
  • Blur [link]
  • Transpose [link]
  • RandomGamma [link]
  • RandomRotate90 [link]
  • MedianBlur [link]
  • GaussianBlur [link]
  • GaussNoise [link]
  • GlassBlur [link]
  • CLAHE [link]
  • ChannelShuffle [link]
  • rgb_to_sepia [link]
  • RandomSnow [link]
  • RandomRain [link]
  • RandomFog [link]
  • RandomSunFlare [link]
  • RandomShadow [link]
  • Solarize [link]
  • Equalize [link]

• kornia.filter.

  • GaussianBlur3d @edgarriba

• kornia.feature.

  • DenseSIFT @ducha-aiki https://github.com/kornia/kornia/issues/462

• kornia.color.

  • lab_to_rgb https://github.com/kornia/kornia/issues/239 @HassanAlsamahi
  • rgb_to_lab https://github.com/kornia/kornia/issues/239 @HassanAlsamahi
  • rgb_to_xyz @ceroytres https://github.com/kornia/kornia/pull/436
  • xyz_to_rgb @ceroytres https://github.com/kornia/kornia/pull/436
  • rgb_to_luv @ceroytres https://github.com/kornia/kornia/pull/442
  • luv_to_ rgb @ceroytres https://github.com/kornia/kornia/pull/442
  • ZCA Whitening https://github.com/kornia/kornia/issues/435 TBD

• Tests

  • setup AWS CI @edgarriba @chadell
  • JIT tests https://github.com/kornia/kornia/pull/261

• Documentation

  • Improve tutorials/examples
  • Start using sphinx-doctest https://github.com/kornia/kornia/issues/422

[shijianjian]

It’d be great if these functions could be implemented first: Solarize, Equalize, Posterize, Sharpness.

Hi @miriaford, those functionalities have been implemented and probably be included in the next release. You may try it out now. Sorry for taking too much time on it.

If possible, we are very happy to see an Autoaugmentation example of using Kornia. 😃

[edgarriba]

@ceroytres go ahead with zca whitening. Make a bit of research about how it should be done, I remember to not go that deep with that 😉