Submitting methodology

[moremo...@gmail.com]

Dear.

Where should we submit our codes and methodology?

[MR Mohebian]

On Wednesday, December 11, 2019 at 6:22:23 PM UTC-6, MR Mohebian wrote:
> Dear.
>
>
> Where should we submit our codes and methodology?

Does anyone know!

[Andreas Turzynski]

Dear organizing committee,

could you please also specify, how we should submit the code. Is the submission of a jupyter notebook file (with explaining markdown) sufficient or do you expect self-running program? If so, could you please define the start and end point.

Thank you in advance

[Eduardo Conde-Sousa]

Hi 

Sorry about my absence the last few days

We are discussing some issues about the submission (both submission deadlines and submission methods) and I can't give a definite answer yet (i'll try next monday or thuesday to came back to you regarding this subject).

The idea of the code submission is for you to provide us with an easy-to-use code/compilation that of course should have a README file explaining how to run it (a "run my code for dummies" style).

Since most software don't accept the mrsx files, you should explain which kind of image input your code accepts (for example tif RGB / tiff  8-bit grayscale / jpeg / ...) for us to convert each slide on that specific format (conversions will be done in slide converter) and feed it. As for the starting and ending point, your code should receive a folder with test images and output a csv file with the classification (again, early next week I'll came back to you explaining all the details). If you want it can export other stuff, but the csv is the only one that will be evaluated.

All the best

Eduardo

Eduardo Conde-Sousa

INEB – Instituto Nacional de Engenharia Biomédica

i3S – Instituto de Investigação e Inovação em Saúde, Universidade do Porto
Rua Alfredo Allen, 208 | 4200-135 Porto, Portugal

Telf: +351 220408815
www.bimage.ineb.up.pt

http://www.ineb.up.pt/

https://www.i3s.up.pt/scientific-platform?v=91
www.i3s.up.pt

--
You received this message because you are subscribed to the Google Groups "HEROHE Grand-Challenge" group.
To unsubscribe from this group and stop receiving emails from it, send an email to herohe-grand-chal...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/herohe-grand-challenge/2a75f49c-6da1-404f-bb31-0d7f035c8c89%40googlegroups.com.

[rqui...@gmail.com]

Hi Eduardo (and everyone)!

Just to be sure, this code we should provide is just for scoring/inference right? So imagine should include serialized trained models or similar?

regarding any possible training (if used), that's not the scope of provided code nor will be need to reproduce?

Related: if there is by some reason any difference between provided outputs and outputs while you reproduce/run the code (ex: lib versions,seeds,etc), which one will prevail?

Thanks!

To unsubscribe from this group and stop receiving emails from it, send an email to herohe-grand-challenge+unsub...@googlegroups.com.

[Eduardo Conde-Sousa]

Hi 

please read bellow:

Just to be sure, this code we should provide is just for scoring/inference right? So imagine should include serialized trained models or similar?

regarding any possible training (if used), that's not the scope of provided code nor will be need to reproduce?

If I understood you well, I believe you're right. I don't need the training process, just what it takes to go from a folder with the test images to a file with classifications. Unless you have a good reason to give us the whole code (starting from the training dataset), I think it is not worth it since the training time is probably big enough to prevent us to run that part (although for the advance of science, it would be excellent if everyone made their code completely open (in some github or similar).

 

Related: if there is by some reason any difference between provided outputs and outputs while you reproduce/run the code (ex: lib versions,seeds,etc), which one will prevail?

To be honest I can't imagine how the lib version can change the output in such a dramatic way. I would imagine that if I have a different version of a given library the code wouldn't run at all, but changing a binary classification seems to much. 

Regarding seeds. Do you use any seed in the classification process? If so, you can hard code it so I can use the same.

In the hypothetical situation of obtaining different results, we would have to analyze the case and talk between the various organizers. I imagine that the soft predictions may change a little bit but the odds of that small change to put two runs in different directions from a given threshold seems to me very low.

Either way, if you feel like this is possible, please refer that in the README file and we will try (if possible) to use your settings.

All the best

Eduardo 

To unsubscribe from this group and stop receiving emails from it, send an email to herohe-grand-chal...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/herohe-grand-challenge/2a75f49c-6da1-404f-bb31-0d7f035c8c89%40googlegroups.com.

--

You received this message because you are subscribed to the Google Groups "HEROHE Grand-Challenge" group.

To unsubscribe from this group and stop receiving emails from it, send an email to herohe-grand-chal...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/herohe-grand-challenge/a013cc68-9e5b-42dd-b04d-fd7380e25e6c%40googlegroups.com.