During testing our solutions, it might happen that a certain hypothesis may result in the degradation of the previous score. In cases like that, would it be possible if we selected the best entry and make the submissions accordingly?Posted by: sayak @ March 10, 2021, 9:30 a.m.
The latest successful submission is the one considered for the final test phase ranking.
It's up to you to decide upon your latest submission.
The codes/executables provided should reproduce the results from your latest submission.
This is not a gambling contest, the chance plays no role in this challenge.
Success!Posted by: Radu @ March 10, 2021, 9:57 a.m.
I hope my intentions were clear.
During an experiment, a hypothesis can fail and I hope we can come to an agreement about that. So, if hypothesis A gave me better results than hypothesis B I would prefer having the hypothesis A results. In my question, I simply asked whether that is possible. So, I don't think it's remotely related to gambling.Posted by: sayak @ March 10, 2021, 10:22 a.m.
> It's up to you to decide upon your latest submission.
Also, the current submission system only takes the latest submission and does not allow us to mark a submission to be the one as per our choice. Will it be different when the test data is released?Posted by: sayak @ March 10, 2021, 10:25 a.m.
You will get no feedback (accuracy) on your results submitted during the test phase.
Therefore, you will not be in the position of picking up an entry with the best score on the test data.
Your latest successful submission will be the one evaluated offline and used for the final test ranking.
You will not be able to select among your submissions, the latest one is final.
The reproducibility codes will be used for checking the solutions on new unseen test images.
I hope this helps.Posted by: Radu @ March 10, 2021, 10:34 a.m.
Yeah now, it's clear. I was thinking otherwise.
Thank you for clarifying.Posted by: sayak @ March 10, 2021, 10:56 a.m.