Awesome
napari-segment-anything
Napari plugin of Segment Anything Model (SAM)
Download the network weights here
This napari plugin was generated with Cookiecutter using @napari's cookiecutter-napari-plugin template.
Installation
You can install napari-segment-anything
via pip:
pip install napari-segment-anything
We recommend installing the latest development version because this package is being developed:
pip install git+https://github.com/jookuma/napari-segment-anything.git
IMPORTANT: napari
won't work if you don't have pyqt5
or pyside2
installed.
Instructions
Opening napari-segment-anything
This software is napari plugin, so once you have napari installed you can open it using the command line:
napari <your image path> -w napari-segment-anything 'Segment Anything'
I noticed that sometimes napari fails to load the plugin widget from the command line, go to step 2 from below to load it.
If you prefer the user interface you need to:
- Drag and drop your image into napari to load it;
- Go to the "Plugins" file menu in the upper right corner and select the "Segment Anything" plugin.
- Follow the instructions below for usage.
IMPORTANT: If you get an error make sure you have pyqt5
or pyside2
installed.
Usage
- Interactions are done on the "SAM points" and "SAM box" layers using the existing functionalities of napari. Only rectangle shapes trigger the network prediction.
- For points supervision, left clicks are positive cues (object) and right clicks are negative (background).
- Press the "Confirm Annot." button (or the "C" key) to propagate the current segmentation mask to the label image.
- Use the napari labels layer features to delete or edit already confirmed labels.
Contributing
Contributions are very welcome. Tests can be run with tox, please ensure the coverage at least stays the same before you submit a pull request.
License
Distributed under the terms of the Apache Software License 2.0 license, "napari-segment-anything" is a free and open source software
Issues
If you encounter any problems, please file an issue along with a detailed description.