Skip That Beat: Augmenting Meter Tracking Models for Underrepresented Time Signatures
2025
Read Skip That Beat: Augmenting Meter Tracking Models for Underrepresented Time Signatures (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-12972.txt.
Description
Giovana Morais, Brian McFee, and Magdalena Fuentes tackle a bias in music tracking software in 2025. Models that find beats and downbeats learn mostly from songs in common time, four beats per bar. They falter on music in other meters, like Brazilian samba in two-four time.
Their fix is simple data surgery: cut beat intervals out of four-four songs to fake two-four and three-four examples. Retrained models found downbeats better in the rare meters while keeping their skill on common time. The trick also lifted performance on a samba data set the models had never seen, a cheap gain from data they already had.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 18 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-12972