Cutting Music Source Separation Some Slakh: A Dataset to Study the Impact of Training Data Quality and Quantity

Abstract

Music source separation performance has greatly improved in recent years with the advent of approaches based on deep learning. Such methods typically require large amounts of labelled training data, which in the case of music consist of mixtures and corresponding instrument stems. However, stems are unavailable for most commercial music, and only limited datasets have so far been released to the public. It can thus be difficult to draw conclusions when comparing various source separation methods, as the difference in performance may stem as much from better data augmentation techniques or training tricks to alleviate the limited availability of training data, as from intrinsically better model architectures and objective functions. In this paper, we present the synthesized Lakh dataset (Slakh) as a new tool for music source separation research. Slakh consists of high-quality renderings of instrumental mixtures and corresponding stems generated from the Lakh MIDI dataset (LMD) using professional-grade sample-based virtual instruments. A first version, Slakh2100, focuses on 2100 songs, resulting in 145 hours of mixtures. While not fully comparable because it is purely instrumental, this dataset contains an order of magnitude more data than MUSDB18, the de facto standard dataset in the field. We show that Slakh can be used to effectively augment existing datasets for musical instrument separation, while opening the door to a wide array of data-intensive music signal analysis tasks.Index Termsmusic source separation, sample-based virtual instruments, synthesis, MIDI

Timbre Comparison in Note Tracking from Onset, Frames and Pitch Estimation

“…The dataset used in this work is the Slakh2100 dataset [5]. The MIDI files are aligned with the audio files and are used as the grund truth picth and duration of the notes.…”

A Comparison of Deep Learning Methods for Timbre Analysis in Polyphonic Automatic Music Transcription

“…We test the f 0 detection and note tracking algorithm (Algorithm 1) in the Slakh2100 dataset [31] which provides MIDI files and their synthesized audio files, and we compare the performance of this approach with the OaF model [4]. It is important to note that the tests have been performed on the same dataset although the OaF model has only been trained with the MAESTRO dataset [17] only for piano transcription.…”

“…For our second method, we train the CNN model for polyphonic transcription with the Slakh2100 dataset [31]. It contains 145 h of mixtures and a total of 2100 automatically mixed tracks and their corresponding MIDI files synthesized.…”

Global Structure-Aware Drum Transcription Based on Self-Attention Mechanisms

“…We used the Slakh2100-split2 (Slakh) [62] and the RWC Popular Music Database (RWC) [63] for evaluation because these datasets include ground-truth beat times. The Slakh dataset contains 2100 musical pieces in which the audio signals were synthesized from the Lakh MIDI dataset [64] using professional-grade virtual instruments, and the RWC dataset contains 100 Japanese popular songs.…”