Open source
Empanada
A napari plugin for deep-learning panoptic segmentation of 2D and 3D electron microscopy. It is aimed at people who segment volumetric data and need the result in a tool they already use.
Who it is for
Research and core-facility teams working with volume electron microscopy who want segmentation, fine-tuning, and export without standing up a separate application.
What it does
2D and 3D segmentation
Panoptic segmentation on a single plane or through a volume.
Training and fine-tuning
Train a model, or adapt one, from inside the plugin.
Tiling
Split a large image into tiles and merge the tiles back.
Label cleanup
Split, merge, filter, and edit labels.
Export
Export segmentations, and move model files in and out of the plugin.
CPU and GPU
CPU builds ship with the plugin. GPU execution is used when a device is available.
How a volume moves through it
01
Tile
Large images are cut into tiles the model can see.
02
Infer
Run 2D or 3D inference on CPU or GPU.
03
Clean
Correct splits, merges, and small labels.
04
Export
Write the segmentation, or carry the model to another machine.
Stack and delivery
Python, PyTorch, and napari. Tests run in continuous integration across desktop operating systems. The plugin is published as an installable package.
Adoption
The package has more than 100,000 downloads.
Our role
Our team led the plugin's core API, testing pipelines, documentation, and CI/CD.
That is the work we claim. It is evidence that we can ship scientific learning software with an API, a test suite, and a release path.
Shipping a scientific model is a software problem.
If you need the API, the tests, and the release path, start a conversation.
Start a conversation