Output images

Hi all,

The challenge is over and it was very insightful (huge thanks to the organizers). I was wondering if some participants could show the images they got for some (randomly taken from the training database) inputs, so that I could see how different top submissions were from mine*. It's just out of curiosity and I know there probably will be such images in the final paper but I'd love seeing some results before then.

For example, for image Study_1/image_0_BF_z0.ome.tiff (this was the input image), I got this for Nucleus (this was the ground truth) and this for Mitochondria (this was the ground truth). For reference, I used plt.imshow and cmap="gray" for those figures.

* didn't submit because wasn't that great to be fair which is why I was wondering how much better were really good submissions!

Thanks!

Hi - I only had time to put together a baseline submission, which I didn't think was worthy of a paper. However, maybe my results might be helpful.

(Segmentation models pytorch, efficientnet-B7, 512x512 random crops, 4xTTA, fine-tuned model output on tubulin and actin images).

My code is at https://github.com/jfozard/LMC, with one set of output images and the corresponding weights are at https://huggingface.co/JFoz/LMC_baseline/

Thank you very much!

Hi, I've made my algorithm public: lmc_control. I'm not familiar with the grandchallenge platform, so I don't know if you can directly try out the algorithm (i.e. permission, who is paying for running credits etc), but there is at least 1 public example.

I'm writing the README and will make my repo, together with checkpoints, visible soon (It has actually always been public, just hidden in some branch). Here is the repo still getting cleanup: lmc_GenCellPaint

Hello tle1302 , I have some questions to ask you, would you please give me a contact email.

Hi! Sorry for not replying, I found that here are several emails that got sent straight to the junk folder (guessing because of the email naming pattern). I'm unsure which one is yours but here is the general answer: I will try to have the technical report with all details at least on biorxiv in ~1 week (there might be also other results and instructions how to train this for your new dataset). Thank you for showing interest! I hope that this can help.

If anyone has more questions, please ping me again in 1 week (via email, which will be in the technical report). And sorry for spamming this thread.