Commercial-Broadcast Retrieval on LibriSpeech
50.19PrecisionVLAFP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VLAFPParams. (M)=12.2, Train (s / epoch)=773.9, Infer. (ms / seg)=78.22026.03 | 50.19 | 44.06 | 46.93 | |
| NAFPParams. (M)=16.9, Train (s / epoch)=1186.6, Infer. (ms / seg)=22.82026.03 | 44.74 | 33.37 | 38.23 | |
| AMGParams. (M)=4.4, Train (s / epoch)=285.6, Infer. (ms / seg)=6.72026.03 | 22.51 | 28.82 | 25.28 | |
| HuBERTParams. (M)=315.5, Infer. (ms / seg)=12.22026.03 | 4.53 | 15.48 | 7.01 | |
| wav2vec2Params. (M)=94.4, Infer. (ms / seg)=6.42026.03 | 3.88 | 26.09 | 6.75 | |
| ASTParams. (M)=86.6, Infer. (ms / seg)=34.12026.03 | 3.8 | 25.03 | 6.59 |