Table 3 Mini-SegNet: performance with different kernel size, maxpooling size, and upsampling size on Mini-SegNet (all-accuracy: %, train/test)
From: Deep semantic learning for acoustic scene classification
Mini-SegNet | |||
|---|---|---|---|
Kernel size | (3,3) | (2,3) | (1,2) |
Maxpooling size | (2,2) | (2,3) | (1,2) |
Upsampling size | (2,2) | (2,3) | (1,2) |
Train params | 670,282 | 478,218 | 210,186 |
All-accuracy | 83.45/66.46 | 80.49/71.26 | 50.31/47.54 |