The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
A marcação de áudio semissupervisionada (AT) e a detecção de eventos sonoros (SED) fracamente rotuladas tornaram-se significativas em aplicações do mundo real. Um método popular é o aprendizado professor-aluno, fazendo com que os modelos dos alunos aprendam a partir de pseudo-rótulos gerados pelos modelos dos professores a partir de dados não rotulados. Para gerar pseudo-rótulos de alta qualidade, propomos uma estrutura mestre-professor-aluno treinado com uma política de liderança dupla. Nossos experimentos ilustram que nosso modelo supera o modelo de última geração em ambas as tarefas.
Yuzhuo LIU
Chinese Academy of Sciences,University of Chinese Academy of Sciences
Hangting CHEN
Chinese Academy of Sciences,University of Chinese Academy of Sciences
Qingwei ZHAO
Chinese Academy of Sciences,University of Chinese Academy of Sciences
Pengyuan ZHANG
Chinese Academy of Sciences,University of Chinese Academy of Sciences
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Yuzhuo LIU, Hangting CHEN, Qingwei ZHAO, Pengyuan ZHANG, "Master-Teacher-Student: A Weakly Labelled Semi-Supervised Framework for Audio Tagging and Sound Event Detection" in IEICE TRANSACTIONS on Information,
vol. E105-D, no. 4, pp. 828-831, April 2022, doi: 10.1587/transinf.2021EDL8082.
Abstract: Weakly labelled semi-supervised audio tagging (AT) and sound event detection (SED) have become significant in real-world applications. A popular method is teacher-student learning, making student models learn from pseudo-labels generated by teacher models from unlabelled data. To generate high-quality pseudo-labels, we propose a master-teacher-student framework trained with a dual-lead policy. Our experiments illustrate that our model outperforms the state-of-the-art model on both tasks.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDL8082/_p
Copiar
@ARTICLE{e105-d_4_828,
author={Yuzhuo LIU, Hangting CHEN, Qingwei ZHAO, Pengyuan ZHANG, },
journal={IEICE TRANSACTIONS on Information},
title={Master-Teacher-Student: A Weakly Labelled Semi-Supervised Framework for Audio Tagging and Sound Event Detection},
year={2022},
volume={E105-D},
number={4},
pages={828-831},
abstract={Weakly labelled semi-supervised audio tagging (AT) and sound event detection (SED) have become significant in real-world applications. A popular method is teacher-student learning, making student models learn from pseudo-labels generated by teacher models from unlabelled data. To generate high-quality pseudo-labels, we propose a master-teacher-student framework trained with a dual-lead policy. Our experiments illustrate that our model outperforms the state-of-the-art model on both tasks.},
keywords={},
doi={10.1587/transinf.2021EDL8082},
ISSN={1745-1361},
month={April},}
Copiar
TY - JOUR
TI - Master-Teacher-Student: A Weakly Labelled Semi-Supervised Framework for Audio Tagging and Sound Event Detection
T2 - IEICE TRANSACTIONS on Information
SP - 828
EP - 831
AU - Yuzhuo LIU
AU - Hangting CHEN
AU - Qingwei ZHAO
AU - Pengyuan ZHANG
PY - 2022
DO - 10.1587/transinf.2021EDL8082
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E105-D
IS - 4
JA - IEICE TRANSACTIONS on Information
Y1 - April 2022
AB - Weakly labelled semi-supervised audio tagging (AT) and sound event detection (SED) have become significant in real-world applications. A popular method is teacher-student learning, making student models learn from pseudo-labels generated by teacher models from unlabelled data. To generate high-quality pseudo-labels, we propose a master-teacher-student framework trained with a dual-lead policy. Our experiments illustrate that our model outperforms the state-of-the-art model on both tasks.
ER -