Table 2.
Performance comparison when using different configurations. C represents convolutional layer, P and F are pooling layer and fully connected layer respectively
| Convnet1 | Convnet2 | Orientation error | Accuracy rate | ||
|---|---|---|---|---|---|
| Network | Patch size | Network | Patch size | () Mean (Std.Dev.) | (%) |
| CCF | 23*23*3 | CCF | 23*23*3 | 0.35 (0.25) | 91% |
| CCF | 23*23*3 | CCCF | 23*23*3 | 0.35 (0.25) | 90% |
| CCF | 29*29*3 | CCF | 23*23*3 | 0.45 (0.52) | 87% |
| CCF | 17*17*3 | CCF | 23*23*3 | 0.88 (0.62) | 88% |
| CCCF | 23*23*3 | CCF | 23*23*3 | 0.40 (0.45) | 88% |
| CCF | 23*23*3 | CPCF | 23*23*3 | 0.35 (0.25) | 85% |
| CCF | 23*23*3 | CCF | 17*17*3 | 0.35 (0.25) | 89% |
| CCF | 23*23*3 | CCF | 29*29*3 | 0.35 (0.25) | 87% |
| CPCF | 23*23*3 | CCF | 23*23*3 | 0.58 (0.63) | 86% |