Announcing our new and stable release of Segment Anything for Microscopy! We introduce new features and improve the tool to segment microscopy data with a few clicks, like the mitochondria in the video, first segmented automatically and then corrected interactively.

What's new? - Our tools are now implemented as napari plugin - We publish better finetuned models, also available on - These models enable a better method for automatic segmentation that also underlies the new 3D segmentation

Want to learn more? - Documentation at - Code at - Video tutorials at

Thanks to everyone in the group who contributed, all external contributors and especially @AnwaiArchit who implemented many of these features.

@neurosapien

Thanks. Trying to test now. No simple pip install?

No, the installation requires conda, though we recommend to use mamba, which is a much faster version of conda. See

The is amazing, important and wonderful! Congratulations.

Looks very useful. @cppape, question. what is the best method out there to count cells/neurons in 3D confocal volume based on nuclear staining. Can we use this? but our neuronal bodies are very close to each other and so densely packed.

This can in principal be used! But from your description there may be other well suited methods. I would also look into stardist / cellpose / go nuclear: (the latter offers pretrained methods for 3d nuclear data)

Thanks for sharing. Any insights on how does it work for irregular (non circular-like) objects like meshes of astrocytes or microglia?

Irregular shapes should work fine, but very fine-grained or vessel like structures may not be a good fit, especially in 3D.