Researchers have developed a new method for evaluating expressive MIDI piano performances by integrating contextual embeddings from self-supervised symbolic music models like Aria and CLaMP3. This approach addresses limitations of traditional attribute-based metrics by considering inter-note dependencies and perceptual properties. The study found these models can serve as effective perceptual proxies, aligning with human ratings. Additionally, an open-source library called Pereval has been released to support reproducibility and integrate various performance evaluation utilities, including attribute-scoped and deep feature metrics. AI
IMPACT This research could lead to more sophisticated AI-driven music generation and evaluation tools, improving the quality and expressiveness of synthesized music.
RANK_REASON The cluster describes a research paper published on arXiv detailing a new methodology and an accompanying open-source library for evaluating symbolic music performances.
Read on Hugging Face Daily Papers →
- Aria
- arXiv
- CLaMP3
- DagsHub
- Hugging Face
- Kernel Audio Distance
- Pearson product-moment correlation coefficient
- Pereval
- MIDI
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →