Sound Event Detection on DCASE HEAR challenge
89.1Onset FMSBYOL-S/CvT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BYOL-S/CvTWindow size=1 s, Hop size=50 ms2022.06 | 89.1 | 15.2 | — | |
| Hybrid BYOL-S/CvTWindow size=1 s, Hop size=50 ms2022.06 | 88.9 | 15.3 | — | |
| BYOL-S/CvT + OSWindow size=1 s, Hop size=50 ms2022.06 | 86.5 | 18 | — | |
| openSMILE (OS) onlyWindow size=1 s, Hop size=50 ms2022.06 | 85.7 | 19.4 | — | |
| ABT AudioNTTBackbone=AudioNTT2022.09 | 76.1 | 27.4 | — | |
| ABT ViTC-BBackbone=ViTC-B, Patch size=16 x 82022.09 | 72.2 | 27.5 | — | |
| wav2vec2Window size=1 s, Hop size=50 ms2022.06 | 67 | 32 | — | |
| HEAR wav2vec2.0Backbone=wav2vec2.02022.09 | 67 | 32 | — | |
| BYOL-SWindow size=1 s, Hop size=50 ms2022.06 | 65 | 35.6 | — | |
| CREPEWindow size=1 s, Hop size=50 ms2022.06 | 55.2 | 42 | — | |
| HEAR CREPEBackbone=CREPE2022.09 | 55.2 | 42 | — | |
| Hybrid BYOL-SWindow size=1 s, Hop size=50 ms2022.06 | 52.6 | 50.4 | — | |
| BYOL-S++Window size=1 s, Hop size=50 ms2022.06 | 51.2 | 58.3 | — | |
| BYOL-AWindow size=1 s, Hop size=50 ms2022.06 | 49.9 | 50.3 | — | |
| HEAR BYOL-A*Backbone=BYOL-A*2022.09 | 49.9 | 50.3 | — | |
| HEAR CREPEBackbone=CREPE2022.09 | 47.2 | — | — | |
| ABT AudioNTTBackbone=AudioNTT2022.09 | 40.5 | — | 27 | |
| ABT ViTC-BBackbone=ViTC-B, Patch size=16 x 82022.09 | 37.4 | — | -10 | |
| HEAR wav2vec2.0Backbone=wav2vec2.02022.09 | 35.1 | — | — | |
| HEAR BYOL-A*Backbone=BYOL-A*2022.09 | 25.1 | — | — |