Forums

NTIRE 2017 Super-Resolution Challenge - Track 1: Bicubic downscaling - x2 Forum

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

> Use of external training data and non example-based methods

I have two questions regarding the rules of the competition.

Can code submissions include models trainned on data other than the div2k trainning set?
Are non example-based methods (e.g. self similarity based) eligible for this competition?

Best regards,
Cristóvão.

Posted by: cristovao.a.cruz @ March 15, 2017, 2:45 p.m.

Dear Cristóvão,

1) The code submissions can include models trained on data other than the provided DIV2K training data.
When using extra data the "Extra Data [1] / No Extra Data [0] " flag should be set to 1 in the readme.txt file.

2) Non example-based methods (like self similarity based or interpolation based) are welcomed to the competition.

Best,
-Radu

Posted by: Radu @ March 15, 2017, 4:26 p.m.
Post in this thread