ilastik is a free, open-source toolkit for segmenting and classifying images with machine learning. You draw a few labels with a brush, and the program predicts the rest and shows the result straight away, on 2D images and on 3D volumes. The ilastik download is free; the core program is licensed under the GNU GPL version 2 or later, and the project publishes Windows, macOS and Linux builds.
ilastik comes from biological imaging. Its documentation talks about cells and experimental data, and it has no dental models or clinical workflow. On a dental imaging site it is relevant to people segmenting research or teaching volumes, not to anyone planning treatment on a patient scan.
ilastik at a glance
| Item | Details |
|---|---|
| Latest stable version | 1.4.2 (27 May 2026), checked 9 November 2026 |
| Beta | 1.5.0b1 (3 November 2026), not covered on this page |
| Licence | Free and open source, GPL-2.0 or later for ilastik, LGPL-2.1 or later for its lazyflow and volumina packages |
| Platforms | Windows 10 and up (64-bit), macOS, Linux (64-bit) |
| Windows installer size | About 500 MB; a GPU build is about 2.4 GB |
| Memory | At least 8 GB RAM recommended; 32 GB for Autocontext on large 3D data |
| Official website | ilastik.org |
What it does
- Pixel Classification: train on brush strokes and get class probabilities for each pixel.
- Object Classification: assign segmented objects to classes.
- Carving: extract objects from a 3D volume semi-automatically.
- Density Counting and Tracking workflows for repeated structures and for objects in 2D+3D sequences.
- Batch and headless operation, so a trained project can run on many files.
Who should use it, and who should not
A good fit: a researcher who needs to separate structures in a volume and does not want to write machine learning code.
Not a fit: clinical tooth or nerve segmentation for treatment. For a free scientific viewer with Python scripting, see napari; ilastik is the better choice when segmentation itself is the job.
Installation notes
- Download the regular Windows 64-bit build from the official download page. Choose the GPU build only if you have a compatible NVIDIA card and accept the NVIDIA CUDA licence.
- Run the installer and pick a workflow on the startup screen.
- Prepare your data in a format the documentation lists, such as TIFF, PNG or HDF5.
Limits and what we have not checked
Not tested yet: we have not installed or run this version. DICOM is not in the file format list on the documentation page we checked, so volumes from a CBCT scanner would need converting first. ilastik needs a 64-bit machine. Facts were checked on 9 November 2026 on ilastik.org.
FAQ
Is ilastik free?
Yes. The ilastik program is under the GPL version 2 or later. The GPU-enabled builds also bring NVIDIA’s CUDA licence terms.
Do I need machine learning experience?
No. The project says that you draw labels and see the result, with no machine learning expertise required.
Does ilastik read DICOM files?
DICOM was not in the format list we checked. Convert the data to a listed format such as TIFF or HDF5.
Is ilastik approved for clinical use?
We found no such statement. It is a research tool from biological imaging.
Last checked against official sources: 9 November 2026 (developer website: ilastik.org). Versions and licence terms change — confirm on the developer's site before deploying in a clinic.
