Table 5. Image-wise accuracy (%) using different voting rules (2 and 4 classes).
Classif. | Vote | 4 Classes | 2 Classes | ||||
---|---|---|---|---|---|---|---|
Init. | Exten. | Overall | Init. | Exten. | Overall | ||
CNN | Maj. | 80.0 | 75.0 | 77.8 | 80.0 | 81.3 | 80.6 |
Max. | 80.0 | 62.5 | 72.2 | 80.0 | 75.0 | 77.8 | |
Sum | 80.0 | 68.8 | 75.0 | 80.0 | 75.0 | 77.8 | |
CNN+SVM | Maj. | 85.0 | 68.8 | 77.8 | 90.0 | 75.0 | 83.3 |
Max. | 80.0 | 62.5 | 72.2 | 80.0 | 75.0 | 77.8 | |
Sum | 85.0 | 68.8 | 77.8 | 90.0 | 75.0 | 83.3 |