Adding self-supervised objectives to a supervised multi-pitch estimator improves closed-set performance but triggers degeneration to blank predictions on additional, unlabeled data.
3.1 Training & Evaluation Details We train and validate the model in each experiment on URMP [18] following the splits proposed in [3]
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Investigating an Overfitting and Degeneration Phenomenon in Self-Supervised Multi-Pitch Estimation
Adding self-supervised objectives to a supervised multi-pitch estimator improves closed-set performance but triggers degeneration to blank predictions on additional, unlabeled data.