Skip to content

Pre-trained TensorFlow.js models for brain segmentation, used by NiiVue Desktop

Notifications You must be signed in to change notification settings

niivue/brainchop-models

Repository files navigation

brainchop-models

Pre-trained TensorFlow.js models for brain segmentation, used by NiiVue Desktop.

These models originate from the brainchop project by neuroneural.

Models

Model Description
brain-extract-full Full brain extraction (skull stripping)
brain-extract-light Lightweight brain extraction
parcellation-104 Brain parcellation into 104 regions
parcellation-50 Brain parcellation into 50 regions
tissue-seg-full Full tissue segmentation (GM/WM/CSF)
tissue-seg-light Lightweight tissue segmentation

File Structure

Each model directory contains:

  • model.json — TensorFlow.js model topology
  • model.bin or group1-shard1of1.bin — Model weights
  • labels.json — Label definitions for segmentation classes
  • settings.json — Model configuration (input shape, preprocessing, etc.)
  • preview.png — Preview image of model output
  • colormap.json / colormap3.json — Color map for visualization (where applicable)

Usage

These files are downloaded on demand by niivue-desktop during its build process via scripts/download-brainchop-models.js. They are fetched from:

https://raw.githubusercontent.com/niivue/brainchop-models/main/<model-name>/<file>

License

The original models are from the brainchop project. Please refer to that repository for model licensing details.

About

Pre-trained TensorFlow.js models for brain segmentation, used by NiiVue Desktop

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published