Table 6. The classification accuracy (%) for the 500 DCT features extracted from different combination of CNNs used in Gastro-CADx using Dataset I.
| Features | Linear | Quadratic | Cubic |
|---|---|---|---|
| ResNet+DarkNet | 95 | 95 | 95.3 |
| AlexNet+DarkNet | 95 | 95 | 95.1 |
| DenseNet+DarkNet | 47.7 | 47.5 | 46.6 |
| AlexNet+ResNet | 94.2 | 94.2 | 94.4 |
| AlexNet+DenseNet | 95.5 | 95.7 | 96.1 |
| ResNet+DenseNet | 96.2 | 96.2 | 96.3 |
| AlexNet+ResNet+DarkNet | 95.5 | 95.8 | 95.9 |
| AlexNet+ DenseNet +DarkNet | 95.8 | 95.9 | 96 |
| ResNet +DenseNet+DarkNet | 96.3 | 96.5 | 96.5 |
| AlexNet+ResNet+DenseNet | 96.8 | 97 | 97.1 |
| AlexNet+ResNet+DenseNet+DarkNet | 97.1 | 97.3 | 97.3 |