<p>May I clarify what is meant by “number of videos that were processed from the test set is lower than the number in the full test set”?</p>
ChallengeSinga
2021-10-13 17:10
0
<p>I think he meant by: suppose if the test set has 1000 videos and if the submission can process only 40 test videos</p>
<p>How the AUROC values will be computed in such a case?</p>
None
aisg_ben
reply
ChallengeSinga
2021-10-14 00:11
0
<p>May I clarify that by “the submission can process only 40 test videos”, you meant that the submission.csv output file would only have 40 probabilities?</p>
ChallengeSinga
reply
aisg_ben
2021-10-18 22:08
0
<p>Yes </p>
<p>or it would contain all 1000 probabilities (filled with 0.5 or random for the unprocessed videos)?</p>
None
aisg_ben
reply
ChallengeSinga
2021-10-19 00:35
0
<p>If some videos do not have a corresponding probability, then there would be a file length error. In other words, each video must have a probability of it being fake. The AUC score is then calculated over the entire test set.</p>
None
aisg_ben
ChallengeSinga
None
aisg_ben reply ChallengeSinga
ChallengeSinga reply aisg_ben
None
aisg_ben reply ChallengeSinga