Skip to content

← All apps

StarDist2D

Segment star-convex nuclei and cells in 2D images with StarDist, from a notebook, Fiji, napari or QuPath.

SegmentationExperimentalnapari plugin: experimentalFiji plugin: experimentalQuPath plugin: experimental

Stardist2DDocumentation

About

A LabConstrictor port of the ZeroCostDL4Mic StarDist 2D notebook. Train a StarDist 2D model on your own images and masks, check its quality against ground truth, export it to the BioImage Model Zoo, and predict on new images, locally or on Google Colab.

The same installed environment also gives Fiji, napari and QuPath a StarDist 2D segmentation tool: choose an open image and a pretrained or your own trained model, and get a label image, outlines (ROI Manager, Shapes layer or QuPath annotations) and a table of per-object measurements.

Please cite StarDist when you use this app: Schmidt U., Weigert M., Broaddus C., Myers G. Cell Detection with Star-Convex Polygons. MICCAI 2018, Lecture Notes in Computer Science 11071, pp. 265-273 (doi:10.1007/978-3-030-00934-2_30). Please also cite ZeroCostDL4Mic, the notebook this app is built on: von Chamier L., Laine R. F. et al. Democratising deep learning for microscopy with ZeroCostDL4Mic. Nature Communications 12, 2276 (2021) (doi:10.1038/s41467-021-22518-0), and LabConstrictor, which packages it (doi:10.48550/arXiv.2603.10704).

Works with

The authors say it works with:

  • 2D images as TIFF: single-channel fluorescence or RGB (H&E); stacks are predicted frame by frame in the notebook
  • Training: matching folders of images and label masks (same file names, TIFF)
  • A pretrained model (2D_versatile_fluo, 2D_versatile_he, 2D_paper_dsb2018) or your own StarDist 2D model folder

Produces

The authors say it produces:

  • Label images (one integer per object), TIFF
  • Object outlines: ImageJ ROI zip, GeoJSON (Fiji ROI Manager, napari Shapes, QuPath annotations)
  • Per-object table (centroid, area, probability) and object counts, CSV
  • Trained model folder, quality-control metrics (IoU, F1, panoptic quality) and PDF reports

Get it on your computer

Install the app on your own computer: download the installer for your system, and the steps for that system open below the buttons. You do this once; installing needs an internet connection, because the installer downloads the app's Python packages.

v1.2.0 · 14 November 2026 · files come from github.com/CellMigrationLab/Stardist2D/releases

Windows

Windows installer322.7 MB · .exe

Installers built with LabConstrictor are not code-signed yet, so Windows may warn you. Run files only from this app's GitHub release. See the Windows steps.

macOS

Apple Silicon (M1 or later)198.4 MB · .pkg

Intel Macs: not supported by this app's authors.

Not sure which? Open the Apple menu, then About This Mac: Intel means Intel, M1 or later means Apple Silicon. If macOS refuses to open it, see the macOS steps.

Linux

Linux installer232.9 MB · .sh

Needs a terminal. Run it with: bash <file name>. See the Linux steps.

How to install on Windows
  1. Download the .exe file and double-click it.

  2. If Windows shows "Windows protected your PC", click "More info" and then "Run anyway". The installers are not code-signed yet.

    Windows protected your PC: click More info, then Run anyway

  3. Follow the on-screen prompts. Launch the app from the Start Menu or your desktop.

How to install on macOS
  1. Choose your file: open the Apple menu, then About This Mac. If the Chip or Processor says Intel, download the Intel installer; if it says M1, M2 or similar, download the Apple Silicon one.
  2. Double-click the .pkg file. If macOS refuses to open it, go to System Settings, Privacy & Security, scroll down to the message about the package and click "Open Anyway" (it appears only after you tried to open the file once). If you do not see it, try opening the file again, then check this page again.
  3. Follow the prompts (we recommend "Install only for me"). Launch the app from your Applications folder.
How to install on Linux
  1. Download the .sh file.
  2. Open a terminal in the download folder and run bash followed by the name of the file you downloaded, for example: bash MyApp-0.1.0-Linux-x86_64.sh
  3. Follow the prompts, then launch the app from your applications menu.

All systems, with pictures: Install an app.

Notebooks to open in your browser

These run on Google's servers in Google Colab: you need an internet connection and a Google account, and any data you load into the notebook is handled by Google. The desktop app runs on your computer.

Using it

  1. Fiji, napari or QuPath: open an image, start LabConstrictor Tools and choose StarDist 2D: Segment nuclei
  2. Pick 2D_versatile_fluo for fluorescence or 2D_versatile_he for H&E, keep the default thresholds, and run
  3. Notebook: open the StarDist 2D notebook from the app (or Colab) and follow sections 1 to 6; section 3.1 gives example data offline
  4. To train your own model: put images and masks in two folders with matching names, then run sections 3 and 4

For any LabConstrictor desktop app: after installing, open the app from the Start menu (Windows), the Applications folder (macOS) or your applications menu (Linux). A terminal window opens with it: keep it open while you work. Its Welcome notebook lists the app's notebooks; the notebook usage guide shows how to work with it.

Read more: Documentation.

Limitations

The authors say:

  • 2D only; use the StarDist 3D app for volumes
  • Objects must be roughly star-convex (round nuclei, cells); long or branched shapes segment poorly
  • GPU training only on Linux with an NVIDIA GPU; Windows and macOS train on the CPU (slow), prediction works everywhere
  • Pretrained models are downloaded on first use (internet needed once)
  • QuPath receives outlines as annotations, not detections

How to cite

The authors give this publication to cite: doi:10.1007/978-3-030-00934-2_30

Read the paper

Name the version you used (latest release: v1.2.0 · 14 November 2026).

All downloads

Every installer of version 1.2.0 (5 files, with checksums)

Released 14 November 2026. Files come straight from the app's GitHub release.

Installers of the latest release
SystemFormatSizeFileSHA-256 (from GitHub)
Linux 64-bit (x86_64).sh232.9 MBStarDist2D-1.2.0-Linux-x86_64.sh616ddedc14fb923272a57f238c7eff4da29f2c1440763f37279a5e401f0d8f08
macOS (Apple Silicon).pkg198.4 MBStarDist2D-1.2.0-MacOSX-arm64.pkgdfd7b3d491dc07e8383875d20e53e2458f6af08431f2511f66f045f336f73e26
macOS (Apple Silicon).sh198.7 MBStarDist2D-1.2.0-MacOSX-arm64.sh6de43965eefc4a4aac44e25ee046a68a1ceba1d97558b2fb66fc8d131f843ec7
Windows 64-bit (Intel/AMD, x86_64).exe322.7 MBStarDist2D-1.2.0-Windows-x86_64.exec36c7d83c6795c4f47d4c641681aaa08ebbe12ed650dfc6ff3c9cd2246b68365
Windows 64-bit (Intel/AMD, x86_64).msi321.8 MBStarDist2D-1.2.0-Windows-x86_64.msi8afe7e379eef9630064c5eee959950f27bacb3d42258e4a8d65cb75eeaca732e

Details

Kind
Segmentation
Development status
Experimental
Latest release
v1.2.0 · 14 November 2026
License
MIT
Repository owner
CellMigrationLab
Also in (plugins)
napari plugin: experimental Fiji plugin: experimental QuPath plugin: experimental
Topics
segmentation, nuclei, StarDist, deep learning, ZeroCostDL4Mic
Links
Report a problem · Documentation

Edit this page on GitHub