FAQ

Most asked questions in December

FAQ

In the past months we have received some questions about specific aspects of the challenge. In case you would also find our responses/clarifications helpful, here they are…

  • The SPEAR challenge does not require people to submit their model nor does it prevent you from submitting your results and model in any future journal/conference article. The challenge will only ask you to provide a brief description of your method to help everyone understand the approach that you took. More details on the submission will be provided soon.
  • Processing at a lower sampling frequency than the provided 48kHz is of course allowed but submitted results will have to be upsampled to the required 48kHz sampling rate. More details on the submission will be provided soon.
  • Some people highlighted that the Cone of Silence pre-trained model has been trained on a different microphone array geometry. While we agree that this leads to impractical use of it for SPEAR, it may still be useful and does not break the challenge’s rules.
  • Some proposed pre-trained models were rejected as they were trained on various versions of the Wall Street Journal corpus. Since this dataset is not readily available to everyone, it does not fit the requirement of the challenge.
  • Some issues in installing the conda environment have been raised and addressed in this GitHub thread. We thank Dr. Aditya Arie NUGRAHA for his help in this matter.

Bonus: December survey

End of the year celebrations are around the corner and after that the release of the SPEAR evaluation set. But before this last stretch, we would appreciate if you could answer a few questions to help us plan.

Thanks to everyone who already answered. If you have not done so yet, you can do it here, it only takes 2min. Thanks in advance!

More details about the submission, listening test and workshop are coming soon to the website.