Sound Event Detection on DCASE challenge task 4 2023 (test)
0.587PSDS1ATST-SED
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ATST-SEDExtra training data (AudioSet clips)=true2023.09 | 0.587 | 0.812 | |
| ATST-SEDExtra training data (AudioSet clips)=false2023.09 | 0.583 | 0.81 | |
| PaSST-SEDExtra training data (AudioSet clips)=false2023.09 | 0.555 | 0.791 | |
| MFDConv-BEATSExtra training data (AudioSet clips)=true2023.09 | 0.552 | 0.794 | |
| FDY-LKA-BEATSExtra training data (AudioSet clips)=true2023.09 | 0.546 | 0.807 | |
| AST-SEDExtra training data (AudioSet clips)=false2023.09 | 0.514 | — | |
| SE-CRNN-16 + EEFD + TAKD + AFL + DAParam=1.3M, Dual Knowledge Distillation=True, Asymmetrical Focal Loss=True, Data Augmentation=True2024.02 | 0.474 | 0.698 | |
| SE-CRNN-16 + EEFD + TAKDParam=1.3M, Dual Knowledge Distillation=True2024.02 | 0.465 | 0.69 | |
| SE-CRNNParam=4.5M2024.02 | 0.456 | 0.693 | |
| Xiao-FMSG-task4a-1Param=2.8M2024.02 | 0.455 | 0.695 | |
| Li-USTC-task4a-7Param=2M2024.02 | 0.451 | 0.673 | |
| SE-CRNN-8 + EEFD + TAKD + AFL + DAParam=400K, Dual Knowledge Distillation=True, Asymmetrical Focal Loss=True, Data Augmentation=True2024.02 | 0.428 | 0.636 | |
| Barahona-AUDIAS-task4a-1Param=1M2024.02 | 0.427 | 0.625 | |
| SE-CRNN-8 + EEFD + TAKDParam=400K, Dual Knowledge Distillation=True2024.02 | 0.407 | 0.628 | |
| Baseline CRNNParam=1.2M2024.02 | 0.366 | 0.575 |