Martin Weigert

Hi, Thanks for using stardist! We haven't yet implemented multi-gpu training. In theory it should not be too hard to incorporate it though (something along the lines of this [discussion](https://www.kaggle.com/c/tgs-salt-identification-challenge/discussion/66458))....

Hi @pouyadfd , 1) Try to always call `model.export_TF()` at the end of a pipeline (i.e. after `model.optimize_thresholds`) and see whether the error persists 2) Please post a code snippet...

Hi @SarkkinenJ , > Hi all, and a huge thanks for developers of StarDist! In an other project StarDist ensured the succession of the project, when it made segmentation of...

Hi, you can get instantiate the paper model via ```python from stardist.models import StarDist2D model = StarDist2D.from_pretrained('2D_paper_dsb2018') ``` and then predict the masks yourself.

Did you normalize the images before prediction?

Did you follow https://github.com/stardist/stardist/blob/master/examples/2D/3_prediction.ipynb ?

Note that in the paper we used a subset (fluorescence images) of the DSB2018, which you can find here: https://github.com/stardist/stardist/releases/download/0.1.0/dsb2018.zip

Hi @psobolewskiPhD , thanks a lot for reporting/testing this! Yes, it seems that this is a version/platform depending bug, although the [respective thread](https://forum.image.sc/t/issue-with-stardist-w-or-w-o-trackmate-with-large-images/60373/10) in the image.sc forum indicates it affects...

Ok, can confirm that on macOS 12 with tf 2.6 everything is fine. ``` platform: macOS-12.0.1-arm64-arm-64bit tf: 2.6.0 ```

Hi Greg, > In my pull request I propose a solution which produces the same segmentation, but prevents these unproductive candidate comparisons and reduces the total execution time 40 minutes,...