2020 VIPriors Image Classification Challenge Forum

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

> Image urls

Dear organizers,

Thanks so much for organizing this competition.

I could not find ImageNet (fall12) images and/or corresponding urls. I submitted on http://image-net.org two months ago, but it seems they are not processing the submissions at the moment.

Would you be able to provide the image urls of ImageNet (fall12)?

Thanks,
Alisher

Posted by: alisher_ai @ June 11, 2020, noon

Hi Alisher,

Unfortunately we cannot help you in acquiring the ImageNet dataset, as it would not be legal for us to share the dataset. We hope you understand. :)

Best,
Robert-Jan

Posted by: rjbruin @ June 11, 2020, 12:01 p.m.

Hi Robert-Jan,

Thank you very much for your prompt reply :) appreciated.

I see. But, then, there is no way for me to participate in this competition, which is very annoying (that ImageNet webpage is not responding at all). But it would be great if you were able to provide the urls at least. But, I understand the situation.

I know this is not the best place to ask this question, but when submitting a paper to the workshop, should we stick on the datasets provided in the competition? Would it still be okay if we perform experiments and show the results on other datasets as well?

Thank you for your consideration,
Alisher

Posted by: alisher_ai @ June 11, 2020, 12:08 p.m.

Hi Alisher,

I understand it is frustrating. I feel the same way. If we had known obtaining ImageNet would be so difficult we would have considered using another dataset, as of course we'd like as many participants as possible to the challenge and hate to see discrimination based on dataset access. Unfortunately we cannot change datasets in the midst of the competition, so I don't see a legal way out of this situation.

As for your other question: we don't have any specific recommendations about which datasets to use in your experiments. We aim to organize an unbiased peer review process for the paper track, so your paper will not be judged favorably or less favorably based on its usage of our dataset. I cannot tell you what the opinion of our reviewers will be as to how many or which datasets are appropriate for your experiments.

Best,
Robert-Jan

P.S. I also replied to the email you sent yesterday. Did you get that response? If not, we might need to check if our email address works correctly.

Posted by: rjbruin @ June 11, 2020, 12:11 p.m.

Hi Robert-Yan,

Thanks once more for your prompt reply :)

Ohhh, I have somehow missed the email yesterday. Thank you very much.

Posted by: alisher_ai @ June 11, 2020, 12:15 p.m.
Post in this thread