AIMAIM
Research

Fall 2023 – present

Evaluator Sound

An app that helps musicians practice more effectively by listening to their playing. Spectrogram analysis and multi-modal transformers detect extra, missed, and wrong notes and return an annotated score, building on our Polytune error-detection model (AAAI 2025).

Evaluator project illustration
Drawing by Cecilia Ines Sanchez.

Evaluator Sound helps musicians practice more effectively by listening to their playing. It uses spectrogram analysis and multi-modal transformers to tell correct from incorrect playing, detecting extra, missed, and wrong notes, and returns an annotated score so the player knows exactly what to correct.

The error-detection model, Polytune, takes audio as input and outputs an annotated music score. It was published at AAAI 2025.

Publications

Detecting Music Performance Errors with Transformers, AAAI 2025.

Team

  • Benjamin "Ben" Shiue-Hal Chou (Lead)