Is Transfer Learning Necessary for Violin Transcription?

TL;DR

This paper investigates whether training violin transcription models from scratch on medium-scale datasets can match or outperform fine-tuned models based on piano pretraining, emphasizing the importance of instrument-specific data.

Contribution

It demonstrates that training from scratch on a medium-scale violin dataset can achieve competitive or superior performance compared to transfer learning from piano models.

Findings

Models trained from scratch perform as well or better than fine-tuned models.

Instrument-specific training can be effective without transfer learning.

Strong violin AMT is feasible with dedicated data collection and augmentation.

Abstract

Automatic music transcription (AMT) has achieved remarkable progress for instruments such as the piano, largely due to the availability of large-scale, high-quality datasets. In contrast, violin AMT remains underexplored due to limited annotated data. A common approach is to fine-tune pretrained models for other downstream tasks, but the effectiveness of such transfer remains unclear in the presence of timbral and articulatory differences. In this work, we investigate whether training from scratch on a medium-scale violin dataset can match the performance of fine-tuned piano-pretrained models. We adopt a piano transcription architecture without modification and train it on the MOSA dataset, which contains about 30 hours of aligned violin recordings. Our experiments on URMP and Bach10 show that models trained from scratch achieved competitive or even superior performance compared to fine-tuned counterparts. These findings suggest that strong violin AMT is possible without relying on pretrained piano representations, highlighting the importance of instrument-specific data collection and augmentation strategies.