Towards Human-in-the-loop Onset Detection: A Transfer Learning Approach for Maracatu
2025
Read Towards Human-in-the-loop Onset Detection: A Transfer Learning Approach for Maracatu (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-2507-04858.txt.
Description
Antonio Sa Pinto adapts music software to an Afro-Brazilian tradition called Maracatu, in a paper from two thousand twenty-five. The task is to spot the exact start of each drum hit in a recording. Maracatu rhythms are dense and complex, so models trained on Western music often fail on them.
He started from two pre-trained networks, one built to find hit starts and one built to track beats. He then fine-tuned them with just five-second labeled clips for each of five percussion instruments. Scores improved sharply, in the best cases by more than fifty points. The beat-based model worked best for instruments that keep time. The best tuning recipe differed by instrument. The approach needs little labeling, which helps traditions that lack large data sets.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-04858