All:
There seemed to be some confusion with the test data (i.e., the lists). For this reason, things have been simplified.
In short: instead of having multiple files, which was a result of the labeled lists being released after last year's challenge. Thus, there are several lists, each containing some information.
Use the following (single lists with all associated metadata):
- Probes (https://www.dropbox.com/s/cxh61i6s91x01ni/probe_list.csv?dl=0)
- Gallery (https://www.dropbox.com/s/cxh61i6s91x01ni/probe_list.csv?dl=0)
The portals will soon be updated, with the same concept done for the validation. For now, I wanted to share these so everyone has the same information.
NOTE: Please be honest. We previously did 100% blind testing. Even though we still want submissions to be scored via codalab server, the labels are now public (i.e., since they were released last year, it would have been unfair for us to remove some, as many had already download). In any case, the last thing we want is to have suspicious activity (i.e., we know the data well-- we know the samples that people get wrong still, and we know the technologies of the ML (and the fundamentals, at that). Please, please be honest. By this: absolutely no modifying submissions manually. The file should be the output of the program and submitted unaltered. This is why the description is required for every "official" submission (else, the submission would be omitted by default).
More information to come! Including a demo for track III (i.e., this track)!
Thank you, and best of luck!
* Gallery (https://www.dropbox.com/s/8xiahy4sl8qr1y8/gallery_list.csv?dl=0) (I mistakenly shared probes 2x above)
Posted by: jvision @ Aug. 28, 2021, 7:39 p.m.