Explanation

March 24, 2026 ยท View on GitHub

This section is for understanding how nnU-Net works and why specific design choices exist.

Core concepts

  • How nnU-Net works
  • What is different in v2?
  • Intensity normalization
  • Region-based training
  • Learning from sparse annotations with ignore labels

Extensibility and customization

  • Modifying the nnU-Net configurations
  • Extending nnU-Net

Migration

  • TLDR migration guide from nnU-Net v1

Contents

  1. 1Core concepts
  2. 2Extensibility and customization
  3. 3Migration