Why do we need this? 1) resolve inheritance in configurations 2) expose otherwise annoying stuff like getting the label manag

This does not actually load the dataset. It merely creates a dictionary where the keys are training case names and the values are dic

Sets everything outside the mask to 0. CAREFUL! outside is defined as < 0, not =0 (in the Mask)!!!

Transforms a 5D array (b, c, x, y, z) to a 4D array (b, c * x, y, z) by overloading the color channel

Reverts Convert3DTo2DTransform by transforming a 4D array (b, c * x, y, z) back to 5D (b, c, x, y, z)

Downscales data_dict[input_key] according to ds_scales. Each entry in ds_scales specified one deep supervision output and its resolut

Takes data_dict[seg][:, index_in_origin], converts it to one hot encoding and appends it to data_dict[key_target]. Optionally removes

Randomly removes connected components in the specified channel_idx of data_dict[key]. Only considers components smaller than dont_do_

Applies random binary operations (specified by any_of_these) with random ball size (radius is uniformly sampled from interval strel_s

regions are tuple of tuples where each inner tuple holds the class indices that are merged into one region, example: regions=

saves 1.6 GB on Dataset017 3d_lowres

DO NOT APPLY NONLINEARITY IN YOUR NETWORK! target mut be one hot encoded IMPORTANT: We assume use_ignore_label is located in

Weights for CE and Dice do not need to sum to one. You can set whatever you want. :param soft_dice_kwargs: :param ce_kwargs:

Wraps a loss function so that it can be applied to multiple outputs. Forward accepts an arbitrary number of inputs. Each input is exp

used for debugging plans etc

used for debugging plans etc

Return an iterator over the shards.

determines whether sample sample_idx in a minibatch needs to be guaranteed foreground

add model config path or file and update registry

Expects batched images with shape BxCxHxW and float format. This transformation may not exactly match apply_image. apply_image is

Build zero-shot classifier weights by iterating over class names 1 by 1 Args: model: CLIP model instance tokenizer: CLIP tokenize