CPU / GPU configuration - 3D spot detection jupyter notebook

[Boehmin]

This is more of a question since I have issues with my Jupyter Kernel dying when running the jupyter notebook. What minimum CPU (+ GPU?) configuration do you recommend to run the notebook successfully? We have a HPC where I can run the notebook on, but I was testing it at the lower end first (4 core, 26GB RAM - and the kernel also died with 14 core, 90GB RAM).

I was also wondering if multi-channel 3D-spot detection was possible, or whether I needed to split the images into their respective channels first? Very keen to try this. The video and initial starting instructions are fantastic. Thanks for putting so much effort into this.

[BBQuercus]

Hey - we haven't experienced any such issues even on smaller machines so far. It's more common to be an issue if the image is too big. If you let us know what notebook you're using and which images we can try to help out.

For all non-2D applications you'll have to split the image into the respective channels/stacks first. In our internal analysis pipeline we've kept it this way to allow for potential multiple models between channels, before subsequently joining/colocalizing etc. The CLI would allow for predictions on larger image but doesn't join them and only outputs the corresponding channel/stack index.