Reproduce “VioTech: Frame-Level Violin Playing Technique Detection with Multi-Scale Mixture-of-Experts” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ting-Kang Wang, Vincent K.M. Cheung, Nan-Tien Lai, Li Su
VioTech predicts violin playing techniques on a 10 ms frame grid, rather than requiring note boundaries before classification. It combines transcription cues with a frozen MERT encoder and gated temporal experts spanning 20 to 2,560 ms. On the RWC test set, its seven-label macro F1 is 76.90%, with gains for pizzicato and spiccato, but lower bow-change F1 than no-MoE.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.