# STONE: Self-supervised Tonality Estimator

arXiv: https://arxiv.org/abs/2407.07408 · HTML: https://mlsanity.com/paper/2407.07408/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 10 Jul 2024.

## Built on by (most cited first, 8 of 8)

- [PESTO: Real-Time Pitch Estimation with Self-supervised Transposition-equivariant Objective](https://mlsanity.com/paper/2508.01488/) (2025)
- [Emergent musical properties of a transformer under contrastive self-supervised learning](https://mlsanity.com/paper/2506.23873/) (2025)
- [Translation-Equivariant Self-Supervised Learning for Pitch Estimation with Optimal Transport](https://mlsanity.com/paper/2508.01493/) (2025)
- [MIDI-RAE-JEPA: Hierarchical Representation Learning and Generation for Symbolic Music](https://mlsanity.com/paper/2607.14537/) (2026)
- [Equivariant Music Transformer](https://mlsanity.com/paper/2608.03920/) (2026)
- [What Makes a Good Layer? Assessing the Layer-Wise Intrinsic Properties of Music Foundation Models](https://mlsanity.com/paper/2608.14819/) (2026)
- [CoJEPA: Combining Contrastive Learning and JEPA for Global-Local Music Representations](https://mlsanity.com/paper/2608.30974/) (2026)
- [From Isolated Feature to Orbits: Discovering Music Concepts via Multi-SAE Alignment](https://mlsanity.com/paper/2610.01864/) (2026)

## Abstract (arXiv)

Although deep neural networks can estimate the key of a musical piece, their supervision incurs a massive annotation effort. Against this shortcoming, we present STONE, the first self-supervised tonality estimator. The architecture behind STONE, named ChromaNet, is a convnet with octave equivalence which outputs a key signature profile (KSP) of 12 structured logits. First, we train ChromaNet to regress artificial pitch transpositions between any two unlabeled musical excerpts from the same audio track, as measured as cross-power spectral density (CPSD) within the circle of fifths (CoF). We observe that this self-supervised pretext task leads KSP to correlate with tonal key signature. Based on this observation, we extend STONE to output a structured KSP of 24 logits, and introduce supervision so as to disambiguate major versus minor keys sharing the same key signature. Applying different amounts of supervision yields semi-supervised and fully supervised tonality estimators: i.e., Semi-TONEs and Sup-TONEs. We evaluate these estimators on FMAK, a new dataset of 5489 real-world musical recordings with expert annotation of 24 major and minor keys. We find that Semi-TONE matches the classification accuracy of Sup-TONE with reduced supervision and outperforms it with equal supervision.