The identity block is the block that has no conv layer at shortcut. # Arguments kernel_size: default 3, the kernel size of mid

The identity block is the block that has no conv layer at shortcut. # Arguments filters: integer, used for first and second conv layers, t

For prediction of test cases just set training=False, this will prevent loading of training data and training batchgenerator initiali

replace genericUNet with the implementation of above for super speeds

- momentum 0.99 - SGD instead of Adam - self.lr_scheduler = None because we do poly_lr - deep supervision = True

- momentum 0.99 - SGD instead of Adam - self.lr_scheduler = None because we do poly_lr - deep supervision = True

changed deep supervision to False :return:

changed deep supervision to False :return:

inference_apply_nonlin to sigmoid

inference_apply_nonlin to sigmoid

Saves a checkpoint every save_ever epochs. :return:

if epoch is not None we overwrite epoch. Else we use epoch = self.epoch + 1 (maybe_update_lr is called in on_epoch_end which is call

if epoch is not None we overwrite epoch. Else we use epoch = self.epoch + 1 (maybe_update_lr is called in on_epoch_end which is call

can be implemented (see nnUNetTrainer) but does not have to. Not implemented here because it imposes stronger assumptions on the pres

Should probably by improved :return:

:param segmentation_export_kwargs: :param model: folder where the model is saved, must contain fold_x subfolders :param list_of_lists: [[

changes best checkpoint pickle nnunettrainer class name to nnUNetTrainer :param experiments: :return:

Pyramid block according to: http://presentations.cocodataset.org/COCO17-Stuff-FAIR.pdf This block generate `M` and `P` blocks.

TP / (TP + FN)

removes all but the largest connected component, individually for each class :param image: :param for_which_classes: can be None. Should

Resizes the images contained in a 4D tensor. # Arguments x: Tensor or variable to resize. height_factor: Positive integer.

also copied ground truth nifti segmentation into the preprocessed folder so that we can use them for evaluation on the cluster

Can be implemented, does not have to :param output_torch: :param target_npy: :return:

due to deep supervision the return value and the reference are now lists of tensors. We only need the full resolution output because

due to deep supervision the return value and the reference are now lists of tensors. We only need the full resolution output because

CAREFUL you need to restore_original_slice_orientation before saving! :param img: :param header: :return:

- we increase roation angle from [-15, 15] to [-30, 30] - scale range is now (0.7, 1.4), was (0.85, 1.25) - we don't do elast

- we increase roation angle from [-15, 15] to [-30, 30] - scale range is now (0.7, 1.4), was (0.85, 1.25) - we don't do elast

we leave out the creation of self.deep_supervision_scales, so it remains None :return:

net_num_pool_op_kernel_sizes is different in resunet