CITATION — REFERENCE ENTRY
reinforcement-learning · henkel2019-rl-score-following
- Citation
- henkel2019-rl-score-following
- Claim ID
- reinforcement-learning
- Assertion
- Henkel and colleagues formulated score following in sheet music images as a multi-modal Markov decision process and trained deep reinforcement learning agents that listen to audio, read score images, and follow along. Trained only on synthesized audio, the agents showed signs of generalizing in first experiments on 16 real piano recordings, which the authors released with note-level alignments.
- Quote
we build on previous work that formulates the score following task as a multi-modal Markov Decision Process (MDP). [...] we design end-to-end multi-modal RL agents that simultaneously learn to listen to music recordings, read the scores from images of sheet music, and follow the music along in the sheet. [...] All experiments so far were based on synthetic data, with audio synthesized directly from the score. We report on first experiments with recordings of 16 real piano performances. [...] The results on this new dataset suggest that our agents are starting to generalize to the real-world scenario even though they were solely trained on synthetic data. We make this set of piano recordings openly available to the research community, along with the ground-truth annotations
- Quote language
- en
- Locator
- section: Abstract, 1
Available in