(1) The final phase of the NTIRE2018 spectral reconstruction challenge is ongoing.
Note that the leaderboard is hidden for this evaluation phase, so you can submit your predictions as usual
(same protocol from the validation/development phase) but the final ranking will be released after the end
of the challenge together with the selected winners and the awards and prizes. The last entry matters.
(2) Each participant/team in the final test phase should fill in a factsheet (solution description and technical details)
The NTIRE2018 factsheet template is available in the data section of the competition:
(3) All the "team leaders" should send an email to firstname.lastname@example.org and email@example.com with the subject:
"NTIRE 2018 Spectral Reconstruction Challenge - (TRACK NAME) (TEAM NAME)" with the filled in factsheet. The body of the email should contain the following:
a) Name of the team that will be used for results submission in the final evaluation phase;
b) Name and email of the team leader;
c) Names of the rest of the team members;
d) Information about the team or any team member affiliation with NTIRE sponsors (Alibaba, NVIDIA, SenseTime, Huawei, CodeOcean, Google, Disney Research, Amazon) or organizers;
e) Name of the team and its users during validation;
f) A download link for their codes / executables. Alternatively,
a link for download of the Codalab dataset of your code / executable or its KEY (as described below).
To upload your code / executable, you can create a dataset in Codalab by using the tab
"My Competitions -> My Datasets", and create a new dataset. After uploading the archived code / executable,
you will have a KEY assigned to your dataset. We need the KEY to access and download your data.
In case of a new version of the code / executable, the team leader will be required to send another email
with the same information for a new dataset and a list of changes.
The reproducibility of the results is a must to be eligible for awards.
For this, each awarded team should provide at least the executables.
Providing also the codes is a BIG plus.
We will have 3 categories of entries in the final test ranking:
1) checked with publicly released codes and executable;
2) checked with publicly released executable;
3) unchecked (with or without released codes or executables).
(4) If a team has more than one solution/method for a challenge track and the solutions are significantly different then it is allowed to have different entries in the final test phase and different associated factsheets/codes.
Note that in this case the team should use different users to submit their final test results in the Codalab system for that particular track.
(5) As mentioned in the "Terms and conditions" there will be several awards. The performance is the main criteria for deciding the winners,
however we plan awards also for a couple of (sub) categories.
The winner(s) and awardee(s) of such a (sub) category will not necessarily be among the overall
top methods of the challenge (in ranked PSNR scores terms).
Novel and interesting methods are encouraged.
All winners/awardees will receive certificates and eventually prizes consisting from money and/or NVIDIA devices, AWS credits.
(6) All the winners/awardees and also the participants with top ranking methods and/or interesting methods
will be invited to co-author the challenge report and to submit papers to the challenge track of NTIRE 2018 workshop.
Thank you for your participation and enjoy the final phase of the challenge!