Under the hood, Dafne completely changed its codebase organization, so that it could be fully installable via pip. I hesitated doing this, because dafne has some complex dependencies that needs to be resolved in a way smarter than pip. For this reason, I have developed the flexidep module to properly manage the installation of tensorflow, SimpleITK, and pyradiomics. I think that the installation is now really easy and professional!
I also acquired a new Mac and packaged Dafne for the M1 CPU. It seems that this was the problem that was making the previous Mac installation file crash on newer systems when running the automatic segmentation. Please try it and let me know!
This release is feature-packed! I am especially proud of the new mask-to-contour conversion framework, which is a bit slow but much more reasonable than the previous one, and not relying on external (binary) libraries! Thanks to this new framework, I could implement slice interpolation, which I think is a very nice tool for manual segmentation.
Unfortunately, I also have bad news. I had applied for a big grant to support Dafne development, but it was not accepted. This means that, for now, I am the only person working on it. I will keep working and developing, but if you would like to contribute, or you know of some funding opportunities, please let me know!
Ah, you will also see an extra feature while running this new release. And if you have read this post, you might have discovered it already. :)
Please bear in mind that this release is still experimental. Contact us to receive your personal API key.