2026:Audio Chord Estimation Results

From MIREX Wiki
Revision as of 13:55, 14 October 2026 by Junyan (talk | contribs) (MIREX 2026 Audio Chord Estimation results)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Submissions

Submission Title PDF Authors
Baseline: Chordino NNLS Chroma v1.1 Link
Baseline: ChordFormer ChordFormer: A Conformer-Based Architecture for Large-Vocabulary Audio Chord Recognition Link
ChordyV-ACE ChordyV: A Hybrid Acoustic-Harmonic Audio Chord Estimation System with Metric Synchronization PDF Mulfie Ansyori Sofyan
MuQ-BTC Audio Chord Estimation Using the SheetSage-Chord Dataset, Pre-trained Features, and Source Separation PDF Marco Viviani, Xinwei Jiang, Zhongyi Huang, Alexandre D'Hooge, Yulun Wu, Yaolong Ju
SLASH SLASH: Audio Chord Estimation with Harmonic Features and Structured Decoding PDF Mehmet Akif Vardar

Test Sets

Main Test Sets

The following datasets are served as pure test sets. No system is allowed to train on them.

  • Billboard 2013: The held-out portion of the McGill Billboard dataset (161 songs), containing mainly western pop songs from the Billboard chart.
  • Yamaha_JPop: A private dataset annotated by Yamaha Corporation (250 JPOP songs).
  • Yamaha_Balanced: A private dataset annotated by Yamaha Corporation (239 songs). While it is still biased towards JPOP songs, the dataset covers a wider range of genres: J.Pop (10.37%), Rock (10.37%), J.Enka (10.37%), J.Kayoukyoku (10.37%), Soundtrack (10.37%), Western Pop (10.37%), Children's Song (10.37%), R&B (6.22%), Hiphop (4.56%), Jazz (2.49%), Dance (2.49%), World (2.07%), Techno (1.24%), Easy listening (1.24%), J.Minyou (1.24%), Others (5.81%).
  • JSIR: to be announced.

Additional Test Sets

These are datasets that may not be strictly held-out test sets. Some models might have been trained on these datasets; for specific details, please refer to the extended abstracts of each model.

  • Billboard 2012: The public portion of the McGill Billboard dataset (188 songs).
  • RWC Popular: 100 pop songs from the RWC (Real World Computing) Music Database. 20% songs with English lyrics and 80% songs with Japanese lyrics.
  • JAAH (new in 2026): 113 jazz recordings of the Jazz Audio-Aligned Harmony dataset, with the official published chord annotations.

Main Results

The following datasets are served as pure test sets. No system is allowed to train on them.

For each metric, the best result is shown in bold. Entries marked * were trained or tuned on that dataset and are not ranked on it.

Billboard2013

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE 54.88 37.58 32.99 25.81 22.16 70.33 78.49 67.62
MuQ-BTC 78.04 74.67 72.37 62.45 60.31 84.50 89.67 82.58
SLASH 82.02 80.37 78.13 62.92 60.90 84.01 86.30 84.37
Baseline: Chordino 71.06 67.18 65.09 48.88 47.06 81.60 83.14 82.71
Baseline: ChordFormer 81.11 80.01 78.98 69.02 68.13 85.22 82.25 92.03

YAMAHA_Balanced

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE
MuQ-BTC
SLASH
Baseline: Chordino
Baseline: ChordFormer

YAMAHA_JPop

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE
MuQ-BTC
SLASH
Baseline: Chordino
Baseline: ChordFormer

JSIR

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE
MuQ-BTC
SLASH
Baseline: Chordino
Baseline: ChordFormer

Additional Results

Below are results on datasets that may not be strictly held-out test sets. Some models might have been trained on these datasets; for specific details, please refer to the extended abstracts of each model.

Billboard2012

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE 57.87 40.63 35.85 28.47 24.68 71.13 79.43 67.60
MuQ-BTC 81.86 79.81 77.39 67.42 64.95 85.98 90.77 83.81
SLASH* 84.54 83.86 81.57 68.86 66.70 85.94 88.87 84.32
Baseline: Chordino 74.04 72.11 70.05 55.24 53.28 83.69 85.33 83.48
Baseline: ChordFormer* 86.29 85.91 84.74 78.50 77.48 89.23 87.00 92.37

* Trained on this dataset: SLASH (public McGill Billboard annotations); Baseline ChordFormer (the 1,217-song set of Humphrey and Bello, which includes the public McGill Billboard songs).

RWC-Popular

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE 63.55 46.76 44.17 34.02 31.99 78.88 80.09 79.23
MuQ-BTC* 88.81 86.65 80.48 75.74 69.55 91.55 94.82 89.64
SLASH* 89.66 89.02 85.15 71.56 68.25 89.43 92.37 86.94
Baseline: Chordino 78.97 77.78 74.13 63.15 59.72 88.64 88.07 89.76
Baseline: ChordFormer* 89.05 87.70 86.17 75.80 74.43 92.21 90.21 94.64

* Trained on this dataset: SLASH (MARL-Chords annotations of the RWC Popular songs); MuQ-BTC (RWC-Pop, 80% of the songs in the submitted fold); Baseline ChordFormer (the 1,217-song set of Humphrey and Bello, which includes RWC Popular).

JAAH

Group MirexRoot MirexMajMin MirexMajMinBass MirexSevenths MirexSeventhsBass MeanSeg UnderSeg OverSeg
ChordyV-ACE 29.37 19.93 15.04 11.21 8.21 68.08 68.62 71.74
MuQ-BTC 58.64 55.74 54.74 43.58 43.26 77.09 85.26 72.65
SLASH* 53.97 51.61 49.17 36.29 34.80 73.61 72.75 78.23
Baseline: Chordino 45.35 41.70 38.81 23.93 21.92 75.61 77.42 76.69
Baseline: ChordFormer 60.39 59.47 57.83 46.10 45.50 75.50 70.10 86.07

* SLASH: decoder settings were tuned on development data that included JAAH songs (the network itself was not trained on JAAH).

Notes

  • MuQ-BTC was trained on SheetSage-Chord, which may overlap with some popular song datasets to a small extent.
Retrieved from "https://music-ir.org/mirex/w/index.php?title=2026:Audio_Chord_Estimation_Results&oldid=15191"