BPM-justering, stems og DAW-eksport

Datasett, versjoner og betingelser holdes atskilt, også kilder uten godkjente rader.

• Beat This! paper, Table 1 ↗

  Installerbare 1 / kilderader 2

  Uten plassering: 1

  En fullstendig offentlig inferenspakke for disse forsøkene er ikke verifisert. Andre vekter med samme modellnavn kan ikke erstatte den.

  Hung

• TelkNet · 2026-09-13 ↗

  Installerbare 220 / kilderader 220

• 2602.19778 · S4.T2 ↗

  Installerbare 2 / kilderader 8

  Uten plassering: 6

  En fullstendig offentlig inferenspakke for disse forsøkene er ikke verifisert. Andre vekter med samme modellnavn kan ikke erstatte den.

  Bortolozzo et al. † · Prior Work (Noisy Student Framework) · Li et al. † · Prior Work (Noisy Student Framework) · 2E1D (SL) (2.2M) · Traditional supervised learning (SL) baseline (trained on full labels with data augmentation) · BTC (SL) (3.03M) · Traditional supervised learning (SL) baseline (trained on full labels with data augmentation) · 2E1D (CL) (50%) · Results of our student models after Stage 2 with Data-Incremental Continual Learning (with clean labels, α = 0.3 ) · BTC (CL) (50%) · Results of our student models after Stage 2 with Data-Incremental Continual Learning (with clean labels, α = 0.3 )

• 2602.19778 · S3.T1 ↗

  Installerbare 1 / kilderader 11

  Uten plassering: 10

  En fullstendig offentlig inferenspakke for disse forsøkene er ikke verifisert. Andre vekter med samme modellnavn kan ikke erstatte den.

  2E1D (2.2M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · FMA (short-form) · 2E1D (2.2M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · DALI (long-form) · 2E1D (2.2M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · MAESTRO+DALI · 2E1D (2.2M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · All (FMA+M+D) · 2E1D (2.2M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · All + KD ( α = 0.3 ) · BTC (3.03M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · FMA (short-form) · BTC (3.03M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · DALI (long-form) · BTC (3.03M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · MAESTRO+DALI · BTC (3.03M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · All (FMA+M+D) · BTC (3.03M) · Our two student models trained only with pseudo-labels (generated by the BTC teacher from unlabeled data) · All + KD ( α = 0.3 )

• 2602.19778 · S5.T3 ↗

  Installerbare 2 / kilderader 8

  Uten plassering: 6

  En fullstendig offentlig inferenspakke for disse forsøkene er ikke verifisert. Andre vekter med samme modellnavn kan ikke erstatte den.

  Count · Bortolozzo et al. † · 2E1D (SL) · BTC (SL) · 2E1D (CL) (50%) · BTC (CL) (50%)

• 2602.19778 · S5.T4 ↗

  Installerbare 0 / kilderader 8

  Uten plassering: 8

  En fullstendig offentlig inferenspakke for disse forsøkene er ikke verifisert. Andre vekter med samme modellnavn kan ikke erstatte den.

  2E1D · 0 · 2E1D · 0.1 · 2E1D · 0.3 · 2E1D · 0.5 · BTC · 0 · BTC · 0.1 · BTC · 0.3 · BTC · 0.4

• Harmony Transformer paper, Table 1 ↗

  Installerbare 0 / kilderader 4

  Uten plassering: 4

  En fullstendig offentlig inferenspakke for disse forsøkene er ikke verifisert. Andre vekter med samme modellnavn kan ikke erstatte den.

  BLSTM · FK · HT · HT*