Sound Event Detection results obtained with the DCASE 2021 SED Baseline system (SED Bs) and the Joint Source Separation + Sound Event Detection proposed methods: Initial model (S0), Two-stage Training (S1, S2), and Joint Training (JT).
收藏NIAID Data Ecosystem2026-05-02 收录
数据链接：
官方服务：
资源简介：
Results are provided over the DESED Public overlap set [20], in terms of PSDS1, PSDS2 and event-based F1 score. Two pre-training methods for Source Separation (FUSS and DESED) and two model selection criteria (Student and Teacher) are compared. The best results for each metric are highlighted in bold.
创建时间：
2024-07-05
