Program
We are very excited to present a program of 1 keynote, 4 invited speakers, and 15 accepted presenters!
Program Summary
Detailed Program
9:10 - 10am Keynote Presentation - Emilia Gómez
Deep Learning for Singing Processing: Achievements, Challenges and Impact on Singers and Listeners
Emilia Gómez1,2, Merlijn Blaauw2, Jordi Bonada2, Pritish Chandna2, Helena Cuesta2
Joint Research Centre (European Commission)1 and Universitat Pompeu Fabra2
This talk summarizes recent advancements on singing voice processing using deep-learning architectures. We discuss on the achievements of current models on a set of singing processing tasks (pitch estimation, source separation and synthesis) in terms of accuracy and sound quality. We then consider the current challenges related to data availability and computing resources. We finally discuss on the impact that these advancements do and will have in singers and listeners when integrated in commercial applications.
11:00 - 12:20am Invited Presentations (20 min each)
Generating Structured Music Through Self-Attention
Anna Huang, Ashish Vaswani, Jakob Uszkoreit, Noam Shazeer, Andrew Dai, Matt Hoffman, Curtis Hawthorne, and Douglas Eck
Google Brain
A “Division of Labour” Approach to Generating Highly Structured and Varied Musical Melodies
Matt McVicar, Gabriele Medeot, Srikanth Cherla, Katerina Kosta, Samer Abdallah, Marco Selvi, Ed Newton-Rex, and Kevin Webster
Jukedeck
Artist Disambiguation in Large Catalogs: Metric Learning and Sampling Strategies
Jimena Royo-Letelier, Romain Hennequin, Viet-Anh Tran, Manuel Moussallam
Deezer
How Stuff Works: LSTM Model of Folk Music Transcriptions
Bob L. Sturm
KTH Royal Institute of Engineering
2:00 - 3:20 Accepted Oral Presentations 1 (20 min each)
Magenta.js: A JavaScript API for Augmenting Creativity with Deep Learning
Adam Roberts, Curtis Hawthorne, and Ian Simon
Google Brain
A Case Study of Deep-Learned Activations via Hand-Crafted Audio Features
Olga Slizovskaia1, Emilia Gómez1,2, and Gloria Haro1
Pompeu Fabra University1, European Commission - Joint Research Centre2
Attention as a Perspective for Learning Tempo-invariant Audio Queries
Matthias Dorfer1, Jan Hajic jr.2, and Gerhard Widmer1
Johannes Kepler University1 and Charles University2
A Hybrid of Deep Audio Feature and i-vector for Artist Recognition
Jiyoung Park1, Donghyun Kim1, Jongpil Lee2, Sangeun Kum2, and Juhan Nam2
NAVER Corp1 and KAIST2
4:30 - 5:10 Accepted Oral Presentations 2 (20 min each)
Singing Style Transfer Using Cycle-Consistent Boundary Equilibrium Generative Adversarial Networks
Cheng-Wei Wu1, Jen-Yu Liu1,2, Yi-Hsuan Yang2, and Jyh-Shing R. Jang1
National Taiwan University1 and Academia Sinica2
Improving DNN-based Music Source Separation using Phase Features
Joachim Muth1, Stefan Uhlich2, Nathanaël Perraudin3, Thomas Kemp2 Fabien Cardinaux2 Yuki Mitsufuji4
École Polytechnique Fédérale de Lausanne (EPFL)1
Sony European Technology Center2
Swiss Data Science Center3
Sony Corporation4
10:00 - 11:00am and 3:30 - 4:30pm Poster Presentations
Block-sparse RNNs Improve Modeling Experimental Evidence from Polyphonic Music Data Sets
Erik Ylipää Hellqvist
Swedish Institute of Computer Science
DeepDrum: An Adaptive Conditional Neural Network for Generating Drum Rhythms
Dimos Makris1, Maximos Kaliakatsos-Papakostas2, and Katia Lida Kermanidis1
Ionian University1 and Institute for Language and Speech Processing, R.C. "Athena''2
Evaluating Repetition Based Melody Prediction Over Different Context Lengths
Radha Manisha Kopparti and Tillman Weyde
City University of London
Explainable Musical Phrase Completion
Gregory W. Johnsen1, Ling Lin1, Lucia Yu2, Andrew Dempsey2, Vishwali Mhasawade2, Daniel Jaroslawicz1, and Iddo Drori1,2
Columbia University1 and New York University2
Harmonic Recomposition using Conditional Autoregressive Modeling
Kyle Kastner, Rithesh Kumar, Tim Cooijmans, and Aaron Courville
Université de Montréal
Incremental Learning for Recognition of Handwritten Mensural Notation
Luisa Micó, José M. Iñesta, and David Rizo
University of Alicante
Modeling Musical Onset Probabilities via Neural Distribution Learning
Jaesung Huh, Egil Martinsson, Adrian Kim, Jung-Woo Ha
NAVER Corp
Recommending Songs to Music Learners Based on Chord Content
Johan Pauwels, György Fazekas, and Mark B. Sandler
Queen Mary University of London
Towards Cover Song Detection Using Siamese Convolutional Neural Networks
Marko Stamenovic
Bose