Audio-to-Text Retrieval on AudioCaps 1K 1.0 (test)
52.4R@1FuseMix(W&H,B)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FuseMix(W&H,B)# (audio, text)=65K, Training Regime=low-data regime2023.12 | 52.4 | 83.4 | 92.4 | |
| ONE-PEACE# (audio, text)=2.4M, Training Regime=internet-scale2023.12 | 51 | 81.9 | 92 | |
| FuseMix(W&H,B)# (audio, text)=50K/15K, Training Regime=low-data regime2023.12 | 50.3 | 81 | 89.6 | |
| LAION-CLAP# (audio, text)=65K, Training Regime=low-data regime2023.12 | 45.3 | 78 | 87.7 | |
| LAION-CLAP# (audio, text)=700K, Training Regime=internet-scale2023.12 | 45 | 76.7 | 88 | |
| CLAP-HTSAT# (audio, text)=70K, Training Regime=low-data regime2023.12 | 41.9 | 73.2 | 84.6 | |
| MMT# (audio, text)=50K/15K, Training Regime=low-data regime2023.12 | 39.6 | 76.8 | 86.7 | |
| ML-ACT# (audio, text)=50K/15K, Training Regime=low-data regime2023.12 | 39.4 | 72 | 83.9 |