Training general-purpose audio tagging networks with noisy labels and iterative self-verification

By Matthias Dorfer, Gerhard Widmer. 2018

28 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Eduardo Fonseca
2. Xavier Favory
3. Xavier Serra
4. Frederic Font
5. Jordi Pons
6. Manoj Plakal
7. Andrés Ferraro
8. D. Ellis

Citations in the corpus

• FSD50K: An Open Dataset of Human-Labeled Sound Events (2020, methodology): “VGG-based architectures have been widely used for both SET [87, 118] and SED [119].”
• Improving Sound Event Classification by Increasing Shift Invariance in Convolutional Neural Networks (2021, result): “This type of architecture has been widely used for SET [2, 22, 23, 24] and is the most competitive baseline for FSD50K when compared to others of higher complexity [16]—which also accords with recent music tagging evaluations [25].”
• Learning Sound Event Classifiers from Web Audio with Noisy Labels (2019, background): “Some attempted to distinguish between the noisy and correct labels with the goal of selecting the latter, for instance, with self-training methods [12, 13].”