Table 3 Performance of frontend and network combinations with DML improvements on the v2 dataset.
From: Dynamic convolution models for cross-frontend keyword spotting
Frontend | Network | Independent | DML-enhanced | DML-improvement | |||
|---|---|---|---|---|---|---|---|
Net1 | Net2 | Net1 | Net2 | Net1 | Net2 | ||
LEAF | TempConvNet | 97.08 | 97.08 | 97.4 | 97.17 | 0.32 | 0.09 |
DynTempConvNet | 97.27 | 97.27 | 97.35 | 97.38 | 0.08 | 0.11 | |
MFCC | TempConvNet | 97.04 | 97.04 | 97.27 | 97.29 | 0.23 | 0.25 |
DynTempConvNet | 97.25 | 97.25 | 97.38 | 97.4 | 0.13 | 0.15 | |
