CLI Overview

March 24, 2026 ยท View on GitHub

This page groups the main nnU-Net v2 command-line entry points by workflow stage.

Dataset preparation

  • nnUNetv2_convert_old_nnUNet_dataset: convert nnU-Net v1 datasets
  • nnUNetv2_convert_MSD_dataset: convert Medical Segmentation Decathlon datasets

Planning and preprocessing

  • nnUNetv2_plan_and_preprocess: run fingerprinting, planning, and preprocessing in one step
  • nnUNetv2_extract_fingerprint: fingerprint only
  • nnUNetv2_plan_experiment: planning only
  • nnUNetv2_preprocess: preprocessing only

Training

  • nnUNetv2_train: train a configuration and fold

Inference

  • nnUNetv2_predict: run prediction using dataset id and stored results
  • nnUNetv2_predict_from_modelfolder: run prediction from an explicit model folder
  • nnUNetv2_ensemble: ensemble multiple prediction folders
  • nnUNetv2_apply_postprocessing: apply postprocessing to prediction outputs

Evaluation and model selection

  • nnUNetv2_find_best_configuration: compare configurations and determine postprocessing
  • nnUNetv2_accumulate_crossval_results: aggregate cross-validation results
  • nnUNetv2_determine_postprocessing: determine postprocessing separately
  • nnUNetv2_evaluate_folder: evaluate a prediction folder
  • nnUNetv2_evaluate_simple: simpler evaluation entry point

Model packaging and sharing

  • nnUNetv2_export_model_to_zip: export a trained model bundle
  • nnUNetv2_install_pretrained_model_from_zip: install a model bundle from zip
  • nnUNetv2_download_pretrained_model_by_url: install a model bundle from a URL

Utilities

  • nnUNetv2_move_plans_between_datasets: reuse plans across datasets
  • nnUNetv2_plot_overlay_pngs: create overlay PNGs for visualization

Help and discovery

All commands support -h:

nnUNetv2_train -h
nnUNetv2_predict -h
nnUNetv2_plan_and_preprocess -h

If you are new to nnU-Net, do not start from this page. Start with: