Dear participants,
The private phase will start tomorrow on October 8th and will last until October 15th, 11:00 AM Israel time (GMT + 3).
During the week of the private phase, each team or competitor will be allowed to submit only 2 submissions, so choose your two final submissions carefully. Again, It is important to notice that only 2 submissions in total in the entire week of the private phase are allowed (not 2 submissions per day - as was in the public phase).*
On October 8th, two new datasets will be available for you in the competition's "DATA" folder in a new folder "Phase 2 - Private":
1) The private test set - which includes 248 unlabeled segments. The private test set will be used for final judging.**
2) A full public test set - This set contains the full tracks of the public test set from which you have received only 106 segments in the public phase. This set contains a total of 284 segments. The full public test set will include the labels (human / animal) of all 284 segments.***
You can use the full, labeled, public test set, in addition to the training set and the auxiliary set, to retrain your model, but you must not use the private test set in any way for training your final model or preliminary analysis. The private test set needs to be used only for making a final prediction.
After submitting your 2 final submissions you won't be able to see the final score until the competition ends and winners announced. candidates for winning will be contacted with more details and requests.
Important Notes:
* Teams, please make sure that you submit only 2 submissions as a team. The automatic mechanism that prevents submitting more than 2 successful submissions per user is not active at the team level. Therefore, in case that the team members will submit more than 2 submissions, only the first 2 submissions will be taken into account, the other submissions will be disqualified.
** The segments in the private test set are selected in a way that guarantees a track has at most a single segment, with the exception of very long tracks that may contribute more than one segment.
The segments in the private test set, the public test set and the training set are not from the same tracks.
*** The segment_id in the full public test set is not the same as it is in the public test set. This change is to guarantee that the segment_id sequence will reflect the order of the segments in each track. A mapping file will be attached (in CSV format) so you will be able to identify the 106 selected segments out of the 284 segments in the full public test set.
Good luck,
MAFAT Challenge Team
yes
by its legs