In order to have a fair challenge, it is imperative that participants do not behave in a way that harms other users or the platform (and thus, indirectly, also other users).
We’ve seen that some participants have uploaded very similar or even equal containers. This is not fair towards other users. We have a limited computational resources and we want everyone to be able to participate.
Also, it seems that some participants / teams are trying to circumvent the “one submission per day” policy by using multiple accounts with the same approach. This is also not fair towards other users.
This challenge is about domain generalization, i.e. to show that your algorithm works on really unseen data. The evaluation on the preliminary test set it there to allow you to check if something went wrong. If you’re trying to iteratively optimize on the test set by getting as many evaluation runs as possible, this defeats the purpose of the challenge.
Please also note that each qualified submission needs to have a description of your algorithmic approach. All qualified submissions to the workshop need to be sufficiently different. If you have different ideas in mind in a larger team and submit different approaches, we welcome this. If you use multiple accounts within a team to circumvent the policy of 1 submission per day, it’s not ok. Please note that abuse of the system might also lead to your account being disabled on the platform.