Pyssl on 3D data (images)

[Nima Mojtahedi]

Great package. I am wondering if there is possibility to use this package for 3D volumetric data (images)?

[Nick Giakoumoglou]

Most self-supervised learning methods typically operate on feature vectors, making it possible to work with 3D data. However, when working with 3D data, two specific modifications are required: 1) applying augmentation transformations directly to the 3D data (instead of 2D) and 2) modifying the feature extractor backbone to accept 3D data as input and produce a 1D latent vector as output.

[Nima Mojtahedi]

Thanks for the comment.
Are you planning to add this feature in the next releases?

[Nick Giakoumoglou]

TBH, I haven't thought about it yet. However, I might consider including it in the near future!

[Nima Mojtahedi]

Cool! Looking forward