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 datasetsnnUNetv2_convert_MSD_dataset: convert Medical Segmentation Decathlon datasets
Planning and preprocessing
nnUNetv2_plan_and_preprocess: run fingerprinting, planning, and preprocessing in one stepnnUNetv2_extract_fingerprint: fingerprint onlynnUNetv2_plan_experiment: planning onlynnUNetv2_preprocess: preprocessing only
Training
nnUNetv2_train: train a configuration and fold
Inference
nnUNetv2_predict: run prediction using dataset id and stored resultsnnUNetv2_predict_from_modelfolder: run prediction from an explicit model foldernnUNetv2_ensemble: ensemble multiple prediction foldersnnUNetv2_apply_postprocessing: apply postprocessing to prediction outputs
Evaluation and model selection
nnUNetv2_find_best_configuration: compare configurations and determine postprocessingnnUNetv2_accumulate_crossval_results: aggregate cross-validation resultsnnUNetv2_determine_postprocessing: determine postprocessing separatelynnUNetv2_evaluate_folder: evaluate a prediction foldernnUNetv2_evaluate_simple: simpler evaluation entry point
Model packaging and sharing
nnUNetv2_export_model_to_zip: export a trained model bundlennUNetv2_install_pretrained_model_from_zip: install a model bundle from zipnnUNetv2_download_pretrained_model_by_url: install a model bundle from a URL
Utilities
nnUNetv2_move_plans_between_datasets: reuse plans across datasetsnnUNetv2_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
Recommended starting path
If you are new to nnU-Net, do not start from this page. Start with: