Skip to main content

Table 5 F-measures for segment-level evaluation on music detection

From: A large TV dataset for speech and music activity detection

  Model Arch. Training data PCEN ORF TV Muspeak OpenBMAT TVSM-test
Third-party method (T1) CNN    0.60 0.93 0.47 0.48
Third-party method (T2) CRNN    0.85 0.99 0.85 0.88
TCN-Cue TCN TVSM-cuesheet   0.79 0.86 0.82 0.88
TCN-P-Cue TCN TVSM-cuesheet 0.86 0.93 0.84 0.90
TCN-P-Pseu TCN TVSM-pseudo 0.87 0.97 0.87 0.93
CRNN-Cue CRNN TVSM-cuesheet   0.89 0.93 0.88 0.93
CRNN-P-Cue CRNN TVSM-cuesheet 0.92 0.94 0.90 0.91
CRNN-P-Pseu CRNN TVSM-pseudo 0.92 0.95 0.91 0.94
  1. The Highest result of each evaluation dataset is marked as boldface