MTG / essentia

C++ library for audio and music analysis, description and synthesis, including Python bindings
http://essentia.upf.edu
GNU Affero General Public License v3.0
2.86k stars 534 forks source link

TempoCNN minimum expected audio length #1387

Open dbogdanov opened 11 months ago

dbogdanov commented 11 months ago

On very short inputs (e.g., 8 seconds) the predictions by TempoCNN are 0 BPM:

[ WARNING ] TempoCNN: On the computation of majority voting, the second candidate, 0, obtained the same number of votes as the winning candidate, 0

The minimum expected length for the the input audio should be clarified in the documentation.

Should the E_WARNING above be more explicit on the reasons for such cases? Or there can be an input length check with a separate E_WARNING (or even exception) before running the network.