data_dict['output_key'] will be a list of segmentations scaled according to ds_scales

returns one hot encodings of the segmentation maps if downsampling has occured (no one hot for highest resolution) downsampled segmentations

this is just a compatibility layer because my target tensor is float and has an extra dimension

ResizeImage layer for 2D inputs. Repeats the rows and columns of the data by factor[0] and factor[1] respectively. # Arguments fac

Residual Encoder, Plain conv decoder

Network has to have NO LINEARITY!

Compute focal loss for multi-class problem. Ignores targets having -1 label

fixes a bug in ConvDropoutNormNonlin where lrelu was used regardless of nonlin. Bad.

fixes a bug in ConvDropoutNormNonlin where lrelu was used regardless of nonlin. Bad.

fixes a bug in ConvDropoutNormNonlin where lrelu was used regardless of nonlin. Bad.

fixes a bug in ConvDropoutNormNonlin where lrelu was used regardless of nonlin. Bad.

Object that holds test and reference segmentations with label information and computes a number of metrics on the two. 'labels' must either be an

patches that are isotropic in the number of voxels (not mm), such as 128x128x128 allow more voxels to be processed at once because we don't h

Combines ExperimentPlannerPoolBasedOnSpacing and ExperimentPlannerTargetSpacingForAnisoAxis We also increase the base_num_features to 32. Th

Same as ExperimentPlanner3D_v21, but designed to fill a RTX2080 ti (11GB) in fp16

Same as ExperimentPlanner3D_v21, but designed to fill a V100 (32GB) in fp16

have 3x conv-in-lrelu per resolution instead of 2 while remaining in the same memory budget This only works with 3d fullres because we use t

preprocesses CT data with the "CT2" normalization. (clip range comes from training set and is the 0.5 and 99.5 percentile of intensities in

attempts to create patches that have an isotropic size (in mm, not voxels) CAREFUL! this one does not support transpose_forward and tran

Preprocesses all data in nonCT mode (this is what we use for MRI per default, but here it is applied to CT images as well)

This preprocessor always uses force_separate_z=False. It does resampling to the target spacing with third order spline for data (just like Ge