Skip to main content

NnUnetParams

AI Segmentation

Configuration parameters for nnU-Net AI model.

Import​

import ScriptingApi as api

# Create parameter instance
params = api.NnUnetParams()

Properties​

models_dir​

Path to the nnU-Net models directory containing trained datasets.

Type: str

dataset​

Dataset name or identifier for the model.

Type: str

configuration​

Model configuration (e.g., '3d_fullres', '3d_lowres', '3d_lowres_high', '3d_cascade_fullres', '2d'). The default is '3d_fullres'. Custom configurations like '3d_lowres_high' (used by TotalSegmentator dental models) are also supported.

Type: str

device​

Processing device ('cpu' or 'gpu'). The default is 'cpu'.

Type: str

folds​

List of model folds to use for ensemble prediction. The default is 0 (single fold).

Type: list

npp​

Number of parallel processes for preprocessing. The default is 0 (sequential mode).

Type: int

nps​

Number of parallel processes for post-processing. The default is 0 (sequential mode).

Type: int

disable_tta​

Disable test-time augmentation for faster inference. The default is false (tta enabled).

Type: bool

trainer​

Trainer class name used during model training. The default is 'nnUNetTrainer'.

Type: str

plans​

Plans identifier for model architecture. The default is 'nnUNetPlans'.

Type: str

checkpoint​

Model checkpoint file to use for inference. The default is the 'checkpoint_final.pth'.

Type: str

See Also​