Creates and configures a contrastive vision-language model. Args: model_name: Name of the model architecture to create. Can be a local mo

:param data: :param seg: :param nonzero_label: this will be written into the segmentation map :return:

per default we use the 50th percentile=median for the target spacing. Higher spacing results in smaller data and thus faster and easi

The default split is a 5 fold CV on all available training cases. nnU-Net will create a split (it is seeded, so always the same) and

function is called dumb because it's dumb

we need too much shit for this function. Problem is that we now have to support region-based training plus multiple input/output formats so there

clean = False will not actually run this. This is just a switch for use with nnUNetv2_plan_and_preprocess where we don't want to rerun finger

Positionally encode points that are not normalized to [0,1].

image can be 2d greyscale or 2d RGB (color channel in last dimension!) Segmentation must be label map of same shape as image (w/o color chan

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

finds all npz files in the given folder and reconstructs the training case names from them

Return set of symbol pairs in a word. Word is represented as tuple of symbols (symbols being variable-length strings).

Get relative positional embeddings according to the relative positions of query and key sizes. Args: q_size (int): size o

Encodes masks to an uncompressed RLE, in the format expected by pycoco tools.

MOVE EVERYTHING INTO THE PLANS. MAXIMUM FLEXIBILITY Ideally I would like to move transpose_forward/backward into the configurations

overwrite_target_spacing ONLY applies to 3d_fullres and 3d_cascade fullres!

Remove padding and upscale masks to the original image size. Arguments: masks (torch.Tensor): Batched masks from the m

IMPORTANT! IF YOU ARE RUNNING THE CASCADE, THE SEGMENTATION FROM THE PREVIOUS STAGE MUST ALREADY BE STACKED ON TOP OF THE IMAGE AS ON

Predicts masks. See 'forward' for more details.

image_properties must only have a 'spacing' key!

Normalize pixel values and pad to a square input.

get dataloader worker seed from pytorch

Removes small disconnected regions and holes in a mask. Returns the mask and an indicator of if the mask has been modified.

separate_z=True will resample with order 0 along z :param data: :param new_shape: :param is_seg: :param axis: :param order:

Compute a binary mask from an uncompressed RLE.

image and segmentation are expected to be 3D (or 1, x, y) selects the slice with the largest amount of fg (how much percent of each class ar

Calculates the image embeddings for the provided image, allowing masks to be predicted with the 'predict' method. Argume

Works for PlainConvUNet, ResidualEncoderUNet

all npz files in this folder belong to the dataset, unpack them all

folder needs the imagesTr, imagesTs and labelsTr subfolders. There also needs to be a dataset.json checks if the expected number of training

Encodes prompts for input to SAM's mask decoder. Arguments: embed_dim (int): The prompts' embedding dimension