OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Voxelmorph

Voxelmorph is an open-source Python library for learning-based image registration and deformation modeling. It provides TensorFlow and PyTorch implementations and documented examples for training registration models, warping images, estimating deformation fields and evaluating segmentation overlap.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Voxelmorph — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Voxelmorph when you need a programmable, local workflow for training or applying learning-based image registration models, and can provide data in the documented formats or adapt its data-loading code.

Who it is for

  • Researchers working on image registration and alignment.
  • Developers building Python workflows for image volumes and deformation modeling.
  • Teams adapting learned registration methods to custom image datasets.

Documented uses

  • Train an unsupervised image-to-image registration network from an image list.
  • Register a moving image to a fixed image or atlas using a trained model.
  • Save a warped image and optionally a predicted deformation field.
  • Evaluate model results with Dice overlap on segmentation labels.
  • Explore atlas construction, annotation warping and deformation-field tutorials.
Before you choose

Requirements & limits

The repository is public and lists an Apache-2.0 license. The README documents installation through PyPI or from a source checkout.

  • The README calls the TensorFlow version stable and directs users to the dev-tensorflow branch; it warns that PyTorch interfaces may change during active development.
  • Training examples accept NIfTI, MGZ and NPZ data; NPZ files use a vol field and may include seg.
  • Example training assumes consistent image shapes, although a customized data generator can handle variation.
  • Registration requires a trained model plus moving and fixed image inputs; the library is not presented as a hosted processing service.
  • The repository includes tutorials for deformable and affine SynthMorph demonstrations, atlas construction, annotation warping and other deformation workflows.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Train and apply a TensorFlow registration model

Train an example image-registration network from a file list, then use its model file to warp a moving image toward a fixed image.

Before you start

  • A TensorFlow checkout using the stable dev-tensorflow branch.
  • An image-list file and training data in NIfTI, MGZ or NPZ format; NPZ files should include vol, with optional seg. Training image shapes are assumed consistent unless the generator is customized.
  • A trained model file and moving and fixed images for registration.

Steps

  1. Prepare the image list and training data in a supported format.

  2. Run the documented TensorFlow training script with the image-list path, model output directory and GPU option.

  3. Pass the resulting model file, moving image and fixed image to register.py, specifying the warped-image output path.

  4. Add --save-warp if the predicted deformation field is also needed.

Expected outcome

The training script saves model weights to the specified model directory; the registration script saves a warped image and can also save the deformation field.

Guide limitations

  • Training data-loading code may need customization for dataset formats; consistent image shapes are assumed unless handled by a custom generator.
  • The example commands use the TensorFlow scripts and a GPU option; the README warns that PyTorch interfaces may change.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
10 October 2026
Original website title
voxelmorph/voxelmorph · Official homepage title · retained in its source language
How we review sourcesSuggest a correction