MIC-DKFZ
Just a Suggestion!
Feature Request: GUI for nnU-Net (Graphical User Interface)
While discussing nnU-Net with a colleague, we realized that there is currently no graphical user interface (GUI) for the tool. This inspired the idea of developing a GUI to make nnU-Net significantly more accessible — especially for medical researchers, clinicians, and domain experts who may not have programming expertise.
Why a GUI?
Although nnU-Net is incredibly powerful and flexible, it currently relies heavily on command-line usage and scripting. This creates a barrier for non-technical users, even those who work directly with medical imaging and have valuable domain insights.
A well-designed GUI would:
Simplify the model training process — no need to remember complex commands or modify scripts Support drag-and-drop data import — easy dataset setup Enable configuration through form fields and dropdowns — select fold, architecture (2D/3D), and parameters without writing code Provide real-time training feedback — graphs for loss, Dice, validation metrics, etc. Allow easy result exploration — view predicted segmentations overlaid on images Enable one-click export — models, logs, and results can be saved and shared easily Promote broader adoption — medical professionals can train and test models independently
Benefits for the nnU-Net Ecosystem
Improved accessibility: Empower medical researchers and clinicians to use state-of-the-art segmentation without needing a data scientist. Wider adoption: Opens the door to institutions and hospitals where technical staffing is limited. More real-world feedback: Enables broader testing across diverse data, leading to valuable insights for improving nnU-Net. Extensibility: Could serve as a front-end for managing multiple models, pipelines, and datasets in research labs.
Introducing a GUI for nnU-Net would make the tool significantly more inclusive and impactful. It would bridge the gap between technical capabilities and clinical expertise — leading to faster, broader, and more meaningful adoption in real-world medical imaging applications.
