Table 3 Accuracy of 3Dcnn in different kernel sizes
From: Palpation localization of radial artery based on 3-dimensional convolutional neural networks
Kernel | 50 pixel | 40 pixel | 30 pixel | 20 pixel | Dropout |
|---|---|---|---|---|---|
60 × 5 × 5 | 0.63 | 0.39 | 0.28 | 0.15 | Yes |
60 × 3 × 3 | 0.79 | 0.52 | 0.37 | 0.25 | Yes |
60 × 1 × 1 | 0.74 | 0.50 | 0.34 | 0.18 | Yes |
30 × 3 × 3 | 0.75 | 0.54 | 0.38 | 0.17 | Yes |
15 × 3 × 3 | 0.87 | 0.61 | 0.45 | 0.27 | Yes |
15 × 3 × 3 | 0.81 | 0.56 | 0.34 | 0.22 | No |
8 × 3 × 3 | 0.78 | 0.59 | 0.31 | 0.20 | Yes |