# Multitask frame-level learning for few-shot sound event detection

**Authors:** Liang Zou, Genwei Yan, Ruoyu Wang, Jun Du, Meng Lei, Tian Gao, Xin Fang  
**arXiv:** [2403.11091](https://arxiv.org/abs/2403.11091)  
**Date:** 2024-03-17  
**Summary page:** https://www.summarizepaper.com/en/paper/2403.11091v1/

## Key points

- Authors focus on few-shot Sound Event Detection (SED)
- Existing methods rely on segment-level predictions
- Introduction of multitask frame-level SED framework
- Proposal of TimeFilterAug for data augmentation
- Achieved impressive F-score of 63.8%
- Secured 1st rank in few-shot bioacoustic event detection category

## Summary

In their paper titled "Multitask frame-level learning for few-shot sound event detection," authors Liang Zou, Genwei Yan, Ruoyu Wang, Jun Du, Meng Lei, Tian Gao, and Xin Fang focus on the challenging task of few-shot Sound Event Detection (SED). This task involves automatically recognizing and classifying sound events with limited samples. The existing methods in few-shot SED primarily rely on segment-level predictions, which can provide detailed and fine-grained predictions for events of brief duration. However, these methods often face difficulties with prediction truncation due to background noise. To address this issue, the authors introduce an innovative multitask frame-level SED framework. This framework leverages frame-level prediction strategies to overcome the limitations of segment-level predictions. Additionally, they propose TimeFilterAug, a linear timing mask for data augmentation. This augmentation technique enhances the model's robustness and adaptability to diverse acoustic environments. The proposed method achieves an impressive F-score of 63.8%, securing the 1st rank in the few-shot bioacoustic event detection category of the Detection and Classification of Acoustic Scenes and Events Challenge 2023. Overall, this research contributes significantly to advancing the field of sound event detection by introducing a novel approach that improves performance in challenging few-shot scenarios.