About
Table of contents
About
Recent advances in AI (including the versatile transformer neural network architecture, generative pretraining, and multimodal modeling) hint at the potential for advances on longstanding music processing problems as well as new applications. This seminar/project course aims to explore the recent research literature contributing toward AI tools serving musicians. Participants will work in teams to motivate, plan, and implement new projects. Throughout the course, the focus will be on musicians as users of the technology, not general consumers as users. Applications under discussion may include: automatic transcription (converting performance audio to symbolic notation), rendering (converting symbolic notation to visual formats), score-performance synchronization (in real time or not), performance analysis, audio transformations (e.g., source separation for music), and other kinds of music audio analysis (e.g., chord identification, beat/tempo tracking). Discussion of evaluation methodology and dataset development is also encouraged.
Class Meetings
Class meetings will be comprised mainly of (1) seminar-style discussions of papers read by participants on assigned topics, and (2) project updates by student teams.