Presenting Kelp-O-Matic v0.7.0! Please note that this release removes the padding and batch_size parameters from library functions and the CLI, resulting in a backward compatibility break. Users requiring these functions are advised to stick with Kelp-O-Matic version <=0.6.1.
The removal of parameters makes way for the introduction of Hann Window classification in KoM. This enhancement aims to improve outputs by generating predictions on overlapping cropped image sections. The combination of these predictions prioritizes those made on pixels near the center of the cropped section, reducing the occurrence of hard edge artifacts in stitched outputs. For some sample images showing the results of this change, please see https://github.com/HakaiInstitute/kelp-o-matic/pull/78.
What's Changed
* Relax pytorch version for linux by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/67
* Dependency updates by dependabot in https://github.com/HakaiInstitute/kelp-o-matic/pull/68, https://github.com/HakaiInstitute/kelp-o-matic/pull/69, https://github.com/HakaiInstitute/kelp-o-matic/pull/71, https://github.com/HakaiInstitute/kelp-o-matic/pull/72
* Update CITATION.cff by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/73
* Update install instructions for better GPU reliability by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/74
* Fix Linux/GPU PyPI installs by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/70
* Move preprocessing transforms into model definition by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/77
* Implement Hann window classifier to reduce edge artifacts by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/78
* Input validation for path arguments by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/79
* Update docs by tayden in https://github.com/HakaiInstitute/kelp-o-matic/pull/80
**Full Changelog**: https://github.com/HakaiInstitute/kelp-o-matic/compare/v0.6.1...v0.7.0