Hi,
I volunteer for
https://www.remap.org.uk/ and we are trying to make a system so a deaf lady can join in with music-based exercise classes through a haptic interface. I would like to use madmom as I have exhausted many other online or realtime beat detection algorithms running on the microcontrollers I am familiar with. I got everything working based on Ian Charnas' example code - Just want to publicly thank Ian for the example and everyone who made/makes madmom. So far I have it running on a Raspberry pi 4 with 4GB and when I run `top` python (2) is top of the list at 330% cpu! The pi is clearly thermally throttling the whole time madmom is running.
I couldn't get reliable results for the beat detection (running in raspian gui so understand there will be desktop etc competing for resources) so I dropped the frame rate to 60Hz and it seemed very nice. Would this cause an issue? Also, are there suggestions on how to reduce resource usage? For example, do different models have different resource demands? Any other advice for our application would be very gratefully received. I plan to use a GPIO to signal to a microcontroller coprocessor that will then send the beat signal over low latency radio link to a haptic device that the user will hold.
I hope that madmom could help many deaf people engage with group music events using haptics.
Thanks again,
Simon