NTIRE 2024 Bracketing Image Restoration and Enhancement Challenge - Track 1 BracketIRE Task Forum

Go back to competition Back to thread list Post in this thread

> Please, considering the competition test size to the validation size.

As the NTIRE 2024 competition nears its end, the number 3 for the final submission is for the final finishing touch, and for those who participated in this competition, they can see what score the deep learning model they developed is getting compared to others and develop their own. It is a valuable number to choose the truly best model among one model.

However, in the case of this competition, there are many difficulties due to upload time due to the test case size being too large, and as the various regions on Earth and the Internet environment are not always the same, these difficulties are causing many people who participated in this competition to develop deep learning. It seems like we're going into an unfortunate situation where we're wasting the last of our time uploading.

Also, depending on the stability of the data upload environment for each team, I think it is somewhat unfair that those who succeed in uploading are given three chances, while those who send by email are given only one chance to decide.

Since the purpose of the competition was very good and it is an academically meaningful competition, I sincerely hope that it will end without any critical operational issues until the end. Also, as there were many people who participated in the competition and put in a lot of effort, the number of test DBs at a level similar to that of validation Db.Please consider changing it to change size of test DB to validation DB.

Posted by: rvslight @ March 17, 2024, 10:07 p.m.

Dear participants,

Relevant notes have been published previously.
We will not change the requirements and instructions, because we think they are fair. The participants are free to choose to submit to the server or email, and a successful submission to the server is possible.
Moreover, we believe that the test set should be as large as possible to better illustrate the performance of the model. Changing it to the validation set may not be a better choice.

Best wishes,
Organizers

Posted by: zhangzhilu @ March 18, 2024, 2 a.m.
Post in this thread