# FUNDAMENTAL FREQUENCY RANGES (Hz) (Based on testing audio manually)
# Copy and comment out most recent previous freq for reference, when amending
# KICK_RANGE = (50, 60)
# KICK_RANGE = (64, 87) # majority of kick drum fundaemtals either 64.60 or 86.13 HZ, so this is a good baseline
KICK_RANGE = (43.1, 87) # majority of kick drum fundaemtals either 64.60 or 86.13 HZ, so this is a good baseline
# LOW_TOM_RANGE = (90, 110)
# SNARE_RANGE = (120, 240)
LOW_TOM_RANGE = (88, 118) # ie between snare and kick??
SNARE_RANGE = (119, 216) # majority of snares in this range, per my analysis
SNARE_SPECTRAL_CENTROID = (3000, 5000) # majority of snares in this range, per my analysis. Could also be MAX 6000
MID_TOM_RANGE = (217, 350) # these ranges are GUESSWORK, based on snare fundamental freq analysis
HIGH_TOM_RANGE = (351, 450)
mir_eval results for IDMT show kick drum recall is poor across multiple genres. analysis shows this is because the range for KICK_RANGE of DS is currently too low and missing 70% of the hits.
Therefore, amend range # KICK_RANGE = (64, 87) to KICK_RANGE = (43.1, 87) in drumscript.notation_generator.constants.py and re-run IDMT, E2E transcription tests to see if this improves recall, and produce better outputs
mir_evalresults for IDMT show kick drum recall is poor across multiple genres. analysis shows this is because the range for KICK_RANGE of DS is currently too low and missing 70% of the hits.Therefore, amend range # KICK_RANGE = (64, 87) to KICK_RANGE = (43.1, 87) in drumscript.notation_generator.constants.py and re-run IDMT, E2E transcription tests to see if this improves recall, and produce better outputs