MedleyDB: A Multitrack Dataset for Annotation-Intensive MIR Research

By Rachel M. Bittner, Justin Salamon, Mike Tierney, Matthias Mauch, Chris Cannam, Juan Pablo Bello. International Society for Music Information Retrieval Conference, 2014

425 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/e0ea1bb742b4958f5c84ece964ac9e3247d44015

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Multi-Scale multi-band densenets for audio source separation (2017, methodology): “• FNN [11]: The source spectra was estimated by feed forward fully connected DNN trained with an additional dataset (MedleyDB [28]).”
Singing Voice Separation: A Study on Training Data (2019, methodology): “Indeed, public training datasets have been regularly released (MIR-1K [17], MedleyDB [18], DSD100 [19] MUSDB [20]) and used to compare different methods, but they are rather small, and often lack diversity.”
Improving singing voice separation using Deep U-Net and Wave-U-Net with data augmentation (2019, methodology): “1 DA: all Data Augmentation, 2 25% of musdb18 kept for early stopping according to [9], 3 MDB: MedleyDB, [23], for the testing phase, 4 Pitch: pitch shifting, 5 Stretch: time stretching, 6 Env.”
Singing Voice Separation with Deep U-Net Convolutional Networks (2017, methodology): “We evaluate the three models on the standard iKala [5] and MedleyDB dataset [3].”
Crepe: A Convolutional Representation for Pitch Estimation (2018, methodology): “Pitch estimation plays an important role in music signal processing, where monophonic pitch tracking is used as a method to generate pitch annotations for multi-track datasets [1] or as a core component of melody extraction systems [2, 3].”
Adaptive Pooling Operators for Weakly Labeled Sound Event Detection (2018, background): “Our motivation for including this dataset in the evaluation is primarily to study the adaptive behavior of the α parameter, and not to achieve the best possible performance (measured by F1).4 3) MedleyDB: MedleyDB [65] is a collection of 122 multitrack recordings from a variety of musical genres and”
A Software Framework for Musical Data Augmentation (2015, methodology): “For this task, we used the MedleyDB dataset, which consists of 122 tracks, spanning a variety of genres and instrumentation [3].”