-
Notifications
You must be signed in to change notification settings - Fork 215
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Export StarDist model, re-use in FIJI #10
Comments
Hi Romain,
that's great to hear!
The pre-trained model should actually work decently on a variety of fluorescence microscopy data.
I have a basic StarDist Fiji plugin working, but can't promise at the moment when it'll be done. Best, |
Hi Uwe, thank you for your prompt answer!
On my original image, the nuclei are 20-60 pixels large and some are fragmented. So I had a look on the training dataset (I have to admit now that I only looked at the first image) and saw that they were more in the range 10-30 pixels. So I tried to dowsample my image with different value, please look at this scaling_results.pdf with the ROIs obtained for 2 images. I think the best results are obtained with the downsampling 0.75 and 0.5.
Sounds promising! Looking foward for a beta! For the moment I modified your jupyter notebook to export the labels as tifs and I transformed the labels image into ROIs in Fiji so:
could save me 1 step. Both solution are fine for prototyping but I would like to include this awesome segmentation in a workflow to process large images. Ideally our users would do it themselves, and while they are now familiar with ImageJ/Fiji, using jupyter notebooks would still require some trainings. Best, Romain |
Hi Romain,
I think the best results are obtained with the downsampling 0.75 and 0.5.
The provided training data contain nuclei with a variety of sizes. Of course, you could train the model only with specific sizes, thereby making a model for your specific nuclei sizes.
For the moment I modified your jupyter notebook to export the labels as tifs and I transformed the labels image into ROIs in Fiji
This will work for non-overlapping nuclei.
the possibility to export the StarDist projections as ImageJ1 ROIs. The Python code is done, but I haven't made an example yet that documents this feature.
could save me 1 step.
This allows you to export overlapping ROIs.
Both solution are fine for prototyping but I would like to include this awesome segmentation in a workflow to process large images.
What do you mean by large images?
Best,
Uwe
|
please help even for e too.. i need to save induvidual ROIS as induvidual files in stardist |
Hi @bharath5673, please ask question like this in the image.sc forum. Best, |
Hi StarDist Team,
I’m playing with StarDist for a couple of days now and it’s amazing !
I just trained a model with your datasets and the network settings from the notebook and use this model to make some predictions on my own images. The results is (as expected) sensitive to scaling but it amazingly good!
I’m now looking for a solution to use the model in FIJI.
I tried to model.exportTF() as in CSBDeep, but the export function is not implemented (yet ?).
Will you release “soon” or do you plan to make a plugin for FIJI?
Thanks again for this amazing tool!
Cheers,
Romain
The text was updated successfully, but these errors were encountered: