Annotation-free Automatic Music Transcription with Scalable Synthetic Data and Adversarial Domain Confusion
This addresses the challenge of annotation scarcity in music transcription for domains with low resources, though it is incremental as it builds on existing adversarial and synthetic data techniques.
The paper tackled the problem of automatic music transcription in domains with no annotated data by proposing a model that uses scalable synthetic audio for pre-training and adversarial domain confusion with unannotated real audio, achieving competitive performance compared to baseline methods without using any real paired MIDI-audio datasets.
Automatic Music Transcription (AMT) is a vital technology in the field of music information processing. Despite recent enhancements in performance due to machine learning techniques, current methods typically attain high accuracy in domains where abundant annotated data is available. Addressing domains with low or no resources continues to be an unresolved challenge. To tackle this issue, we propose a transcription model that does not require any MIDI-audio paired data through the utilization of scalable synthetic audio for pre-training and adversarial domain confusion using unannotated real audio. In experiments, we evaluate methods under the real-world application scenario where training datasets do not include the MIDI annotation of audio in the target data domain. Our proposed method achieved competitive performance relative to established baseline methods, despite not utilizing any real datasets of paired MIDI-audio. Additionally, ablation studies have provided insights into the scalability of this approach and the forthcoming challenges in the field of AMT research.