Research map / Image generation / Autoregressive and latent generation
Enabling Factorized Piano Music Modeling and Generation with the MAESTRO Dataset
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Train modified Onsets & Frames on MAESTRO with audio augmentation | Piano transcription · MAPS configuration 2 test | Frame F1: 82.02 -> 84.91 | Helped |
| Add audio augmentation (ablation) | Onsets & Frames trained on MAESTRO · MAPS configuration 2 test | Note w/ offset & velocity F1: 48.07 -> 51.77 | Helped |
| Add audio augmentation (ablation) | Onsets & Frames trained on MAESTRO · MAESTRO test | Frame F1: 90.15 -> 89.19 | Hurt |
| Add audio augmentation (ablation) | Onsets & Frames trained on MAESTRO · MAESTRO test | Note w/ offset & velocity F1: 77.54 -> 76.04 | Hurt |
| Train Music Transformer on transcribed MAESTRO audio MIDI instead of ground-truth MIDI | Music Transformer · Respective validation splits | Negative Log-Likelihood: $1.84$ -> $1.72$ | Helped |
| Condition WaveNet on MIDI rather than train unconditioned | WaveNet piano synthesis · MAESTRO training/validation | Training loss: 3.72 -> 3.70 | No clear effect |
| Use inferred MIDI rather than ground-truth MIDI to condition WaveNet | WaveNet piano synthesis · MAESTRO training/validation | Training loss: 3.70 -> 3.84 | Hurt |
| Add an offset detection head | Onsets & Frames transcription model · Piano transcription | Note w/ offset F1 | No clear effect |
Closest papers by meaning
- Aria-MIDI: A Dataset of Piano MIDI Files for Symbolic Music Modeling (ICLR 2025 · similarity 0.89)
- Rubato: Transcribing Piano Music with Timestamps (NeurIPS 2026 · similarity 0.88)
- MIDI-GPT: A Controllable Generative Model for Computer-Assisted Multitrack Music Composition (AAAI 2025 · similarity 0.88)
- Structured Multi-Track Accompaniment Arrangement via Style Prior Modelling (NeurIPS 2024 · similarity 0.88)
- MuPT: A Generative Symbolic Music Pretrained Transformer (ICLR 2025 · similarity 0.87)
- Unifying Symbolic Music Arrangement: Track-Aware Reconstruction and Structured Tokenization (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.