Table 1 Performance of frontend and model combinations.
From: Dynamic convolution models for cross-frontend keyword spotting
Frontend | Model | FLOPs | Params | Accuracy | |
|---|---|---|---|---|---|
v1 (%) | v2 (%) | ||||
LEAF | TempConvNet | 5.48M | 52K | 96.2 | 97.08 |
MFCC | 96.37 | 97.04 | |||
LEAF | DynTempConvNet | 6.11M | 62K | 96.47 | 97.27 |
MFCC | 96.50 | 97.25 | |||
