Lighter generalist models

This release contains the light version of the generalist models trained with the nnunet fold_all option, using the whole training set. We include 2 models:

• model_seg_generalist trained on the full data aggregation (TEM + SEM + BF + CARS)
• model_seg_genralist_bf trained on the single modality aggregation of 3 different BF datasets

These models are lighter and faster than the ensemble version. For more information, see #21.

What's Changed

• OOD Experiment Support by @ArthurBoschet in #20
• Image Utilities Addition by @ArthurBoschet in #19
• Added "fold_all" Support by @ArthurBoschet in #17
• Ensembling Experiment Support Added by @ArthurBoschet in #18

Full Changelog: r20240224...r20240416

Initial release of the Multi-Domain segmentation model

This release contains the generalist model checkpoints (trained on 2024-02-06). Only the best checkpoints are included. Better performance can be achieved by ensembling the 5 folds.