# Addressing Missing Labels in Large-Scale Sound Event Recognition Using a Teacher-Student Framework With Loss Masking

arXiv: https://arxiv.org/abs/2005.00878 · HTML: https://mlsanity.com/paper/2005.00878/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 May 2020.

## Built on by (most cited first, 7 of 7)

- [PSLA: Improving Audio Tagging with Pretraining, Sampling, Labeling, and Aggregation](https://mlsanity.com/paper/2102.01243/) (2021)
- [Vocalsound: A Dataset for Improving Human Vocal Sounds Recognition](https://mlsanity.com/paper/2205.03433/) (2022)
- [Symptom Identification for Interpretable Detection of Multiple Mental Disorders](https://mlsanity.com/paper/2205.11308/) (2022)
- [CMKD: CNN/Transformer-Based Cross-Model Knowledge Distillation for Audio Classification](https://mlsanity.com/paper/2203.06760/) (2022)
- [Self-Supervised Learning from Automatically Separated Sound Scenes](https://mlsanity.com/paper/2105.02132/) (2021)
- [Improving Sound Event Classification by Increasing Shift Invariance in Convolutional Neural Networks](https://mlsanity.com/paper/2107.00623/) (2021)
- [Face: Fast, Accurate and Context-Aware Audio Annotation and Classification](https://mlsanity.com/paper/2303.03666/) (2023)

## Abstract (arXiv)

The study of label noise in sound event recognition has recently gained attention with the advent of larger and noisier datasets. This work addresses the problem of missing labels, one of the big weaknesses of large audio datasets, and one of the most conspicuous issues for AudioSet. We propose a simple and model-agnostic method based on a teacher-student framework with loss masking to first identify the most critical missing label candidates, and then ignore their contribution during the learning process. We find that a simple optimisation of the training label set improves recognition performance without additional computation. We discover that most of the improvement comes from ignoring a critical tiny portion of the missing labels. We also show that the damage done by missing labels is larger as the training set gets smaller, yet it can still be observed even when training with massive amounts of audio. We believe these insights can generalize to other large-scale datasets.