Transcoda: End-to-End Zero-Shot Optical Music Recognition via Data-Centric Synthetic Training

TL;DR

Transcoda introduces a data-centric synthetic training approach with normalization and grammar-based decoding, enabling efficient end-to-end zero-shot optical music recognition that outperforms larger models.

Contribution

It presents a novel synthetic data pipeline, normalization of music encoding, and grammar-based decoding to improve OMR accuracy without large datasets.

Findings

Outperforms state-of-the-art baselines on synthetic benchmark with 18.46% OMR-NED

Reduces error rate on historical Polish scans to 63.97% OMR-NED

Trains a 59M-parameter model in 6 hours on a single GPU

Abstract

Optical Music Recognition (OMR), the task of transcribing sheet music into a structured textual representation, is currently bottlenecked by a lack of large-scale, annotated datasets of real scans. This forces models to rely on either few-shot transfer or synthetic training pipelines that remain overly simplistic. A secondary challenge is encoding non-uniqueness: in the popular Humdrum **kern format for transcribing music, multiple different text encodings can render into the same visual sheet music. This one-to-many mapping creates a harder learning task and introduces high uncertainty during decoding. We propose Transcoda, an OMR system built on (i) an advanced synthetic data generation pipeline, (ii) a normalization of the **kern encoding to enforce a unique normal form and (iii) grammar-based decoding to ensure the syntactic correctness of the output. This approach allows us to train a compact 59M-parameter model in just 6 hours on a single GPU that outperforms billion-parameter baselines. Transcoda achieves the best score among state of the art baselines on a newly curated benchmark of synthetically rendered scores at 18.46% OMR-NED (compared to 43.91% for the next-best system, Legato) and reduces the error rate on historical Polish scans to 63.97% OMR-NED (down from 80.16% for SMT++).