Deep Unsupervised Drum Transcription

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, it learns from a large unlabelled dataset. In DrummerNet, the target drum signal is first passed to a (trainable) transcriber, and a (fixed) synthesizer reconstructs the input signal from 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. In the experiments, DrummerNet performs favorably compared to many recent drum transcription systems, both supervised and unsupervised.