# Deep Unsupervised Drum Transcription

## TL;DR

This paper presents DrummerNet, an unsupervised deep learning system for drum transcription that learns from unlabeled data by minimizing reconstruction error, outperforming many existing methods.

## Contribution

Introduces DrummerNet, a novel unsupervised neural network approach for drum transcription that does not require ground-truth labels and leverages large unlabeled datasets.

## Key findings

- Performs favorably compared to recent supervised and unsupervised systems
- Successfully learns drum transcription without ground-truth annotations
- Demonstrates scalability with large unlabeled datasets

## Abstract

We introduce DrummerNet, a drum transcription system that is trained in an unsupervised manner. DrummerNet does not require any ground-truth transcription and, with the data-scalability of deep neural networks, learns from a large unlabeled dataset. In DrummerNet, the target drum signal is first passed to a (trainable) transcriber, then reconstructed in a (fixed) synthesizer according to the transcription estimate. By training the system to minimize the distance between the input and the output audio signals, the transcriber learns to transcribe without ground truth transcription. Our experiment shows that DrummerNet performs favorably compared to many other recent drum transcription systems, both supervised and unsupervised.