> Validation set labels and training/adapting to unlabelled test data

Hello, thanks for organizing the competition! We have a couple of questions:

(1) will labels be released for the existing validation set after the test set becomes available in the testing phase?
(2) is training a model or adapting a model to the test data (in an unsupervised way of course) allowed for test submissions?

Thank you!

Posted by: shisen @ July 13, 2021, 7:55 p.m.

(1) Yes. We will release the validation set's ground truth file during Phase 2.
(2) Yes. You are allowed to use unsupervised adaptation techniques. Please document the same in the accompanying paper if you do so.

Posted by: ambareesh.r @ July 14, 2021, 1:33 p.m.
Post in this thread