Audio Tagging on Balanced AudioSet 1.0 (test)
0.9638AUCProposed Full Model (PSLA Ensemble)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Proposed Full Model (PSLA Ensemble)#Params=13.6M x 202021.02 | 0.9638 | 0.362 | 2.541 | |
| Proposed 68M Model (PSLA Ensemble)#Params=13.6M x 52021.02 | 0.9602 | 0.3527 | 2.479 | |
| Proposed Single Model (PSLA)#Params=13.6M2021.02 | 0.9534 | 0.3192 | 2.374 | |
| Kong2021.02 | 0.949 | 0.274 | 2.316 | |
| Kumar2021.02 | 0.927 | 0.213 | 2.056 | |
| Wu-best#Params=56M2021.02 | 0.927 | — | 2.056 | |
| Wu-minimal#Params=2.6M2021.02 | 0.916 | — | 1.95 | |
| Baseline (EfficientNet)#Params=13.6M2021.02 | 0.9108 | 0.157 | 1.903 | |
| PANNs#Params=81M2021.02 | 0.905 | 0.278 | 1.852 |