Closed r100gs closed 1 month ago
Not currently. We certainly plan to release more models in the future - and to continue improving the method - although it will be at least a couple of weeks before we can do this.
A major limitation is the availability of open training data. I am aware of this, but it is licensed CC BY-NC-SA 4.0
and this has restrictions that aren't compatible with open-source - and it's not clear whether models trained on this data would have the same restrictions (despite the peer-reviewed publication stating that the data is shared under a difference license).
For a discussion on why 'non-commercial' clauses are problematic, see https://freedomdefined.org/Licenses/NC
In the meantime, you can try the fluorescence model after selecting the input channels to include color-deconvolved channels from the brightfield image. I find that this sometimes works quite well, but not consistently enough to recommend in general - which makes sense, as it was not trained on any brightfield images.
Hello,
the extension is running fine and brightfield nuclei detection is good, but is there a model for nuclei and cell detection in brightfield?
Thank you, Stefan