An OMR model 6x more accurate at 1/30 the size
kobi.music
1 thread
I made a OMR (optical music recognition) model that outperforms the next leading released model (legato) on quartet scans by 6x. It's 28m parameters, with a 2m parameter harness, vs legato which is 100m parameters but with a 900m vision encoder strapped on.
Kudos from a person who's also built an OMR engine - the one at Soundslice, which is in my biased opinion the best one. :)
Thank you! I have respect for soundslice for also furthering the field. Cheers