train-with-disc-only-landmarks-variants

This release includes models trained with only discs landmarks (C2C3, C7T1, T12L1, L5S).

Here's a description of each configuration in the release:

• nnUNetTrainer__nnUNetPlans: Default nnUNet trainer configuration.
• nnUNetTrainer__nnUNetPlans_small: Default nnUNet trainer configuration with small patch size (128x96x96).
• nnUNetTrainer_DASegOrd0_NoMirroring__nnUNetPlans: nnUNet trainer with no mirroring data augmentation and nearest-neighbor segmentation interpolation (Instead of the default linear).
• nnUNetTrainer_DASegOrd0_NoMirroring__nnUNetPlans_small: Same as above but with small patch size (128x96x96).
• nnUNetTrainer_DASegOrd0_NoMirroring__nnUNetResEncUNetLPlans: nnUNet trainer with no mirroring data augmentation and nearest-neighbor segmentation interpolation (Instead of the default linear), using residual encoder plans.
• nnUNetTrainer_DASegOrd0_NoMirroring__nnUNetResEncUNetLPlans_small: Same as above but with small patch size (128x96x96).