- Deep Learning for Music Analysis and Generation, NTU
COMME5070, 2024 Fall
- Official
course website at NTU (access limited): https://cool.ntu.edu.tw/courses/41158
- Slides: https://github.com/affige/DeepMIR/
- Syllabus
- W1.
Introduction to the course
- W2.
Fundamentals of musical audio
- W3.
Music classification and transcription
- W4.
Source separation
- W5. GAN
& Vocoders
- W7.
Fundamentals of symbolic music
- W8.
Symbolic MIDI generation
- W9.
Synthesis and style transfer
- W10.
Singing voice generation
- W11.
Text-to-music generation
- W13.
Differentiable DSP models and automatic mixing
- W14.
Miscellaneous Topics
- List of final projects presented
by the class students
- Zero-shot diffusion-based framework for music inverse
problem
– extensions of this project has been submitted to a
conference for publication at
NeurIPS 2025 Workshop on AI for Music (AI4Music’25): “Zero-shot
geometry-aware diffusion guidance for music restoration”
- Diffusion-based piano MIDI
improvisation
- Classical
MuseMorphose
- Diffusion-based guitar tone
morphing
– extensions of this project has been accepted for publication
at APSIPA ASC’25: “Guitar
Tone Morphing by Diffusion-based Model”
- Diffusion-based audio
morphing
- Transformer-based
text-to-music loop generation
- Singing
duet generation
- Story-to-music
generation
- Guitar jamming
- Singing voice conversion
- Tempo
manipulation
- RL-based
classical piano MIDI generation
- Diffusion-based
emotion-based MIDI generation
- Jazz-to-classical
piano MIDI conversion
- Music
review generation
- Symbolic
music emotion recognition