Talk:Audio Tempo Extraction

The proposal is very well written. Cool!

I hope you don't mind if my questions sound naive, as I don't have much background on this topic:

1. What is the audio format? (PCM, 16 bit ?)

Yes, the audio format will likely be PCM, 16 bit

2. Could this situation get partially credit that the tempi are detected, but in a wrong order (detect secondary as primary, but primary as secondary)?

Yes. From my experience, this is the most common error and the the algorithm should get partial credit.

3. Why does the integer multiple of the tempo matter? Why does it have to be test only if the primary or secondary tempo is correctly identified? My intuition is to identify integer multiple would be easier than to identify the exact tempo.

Yes, it is easier to identify an integer multiple of the tempo than the exact tempo, but this is also a case where the algorithm should get "partial credit".

4. Does this task need any training data? Should the participants use their own training data? In that case, the annotation of the testing data should be of no controversy among participants.

We will provide a few excerpts with annotations attached for practice/training.

Just my two cents, Thank you !