DEMUN: Fast and accurate discovery of music notation in very large collections

2026-06-30

AI Summary

This study addresses the challenge of automatically identifying sparsely distributed and unlabeled music score images within massive collections of non-specialized textual documents, such as textbooks and newspapers, where conventional approaches suffer from low efficiency and existing automated methods struggle to simultaneously achieve extremely low false positive rates and high throughput. To overcome these limitations, this work proposes a lightweight two-stage detection framework that integrates image processing and machine learning techniques, specifically optimized for sparse target retrieval in large-scale document corpora. Evaluated on a dataset of four million scanned pages, the method efficiently detected 1,500 pages containing musical notation with a remarkably low false positive rate of only 0.015%. Extrapolating from these findings, the authors estimate the likely existence of 20,000 to 30,000 previously unrecorded music-related documents in archival collections, substantially enhancing the feasibility and scale of cultural heritage discovery.

Abstract

Much of written musical heritage is preserved and digitised at memory institutions: libraries, museums, and archives. Owing to their collection structures, sheet music tends to be concentrated in large subsets that are defined as collections of music, with corresponding metadata that makes the music findable. However, when studying musical life as opposed to individual works, relevant documents often lie outside of these specialised collections: in textbooks, newspapers, other periodicals, pamphlets, and other documents with extensive circulation. But these documents are typically not catalogued as musical documents, and though there may be a lot of such documents overall, in large library collections, they are still extremely sparse. Manual discovery is thus unfeasible. Automated discovery requires an extremely low false positive rate in order to be useful, and must also operate quickly. We present DEMUN: a two-stage lightweight detector of music notation with a false positive rate of 0.015 %. In the test scenario, 4 million images of a national-scale library were processed, out of which 1,500 pages with music notation were discovered, suggesting the entire collection may contain up to 20-30,000 unmarked documents of musical life.

Problem

• music notation discovery
• large-scale library collections
• sparse musical documents
• automated detection
• false positive rate

Innovation

• music notation detection
• false positive rate
• large-scale document analysis
• lightweight detector
• digital heritage