Research Article
Exploring Deep Learning and Transfer Learning for Colonic Polyp Classification
Table 7
Results from i-Scan1 database with images resized to 224 Ć 224 using the last fully connected layer and the prior fully connected layer.
| ā | CNN-F | CNN-M | CNN-S | CNN-F MCN | CNN-M MCN | CNN-S MCN | Google LeNet | VGG VD16 | VGG VD19 | AlexNet | AlexNet MCN | |
| Prior fully connected layer | 89.33 | 90.67 | 90.00 | 82.00 | 90.67 | 91.42 | 90.67 | 85.33 | 82.67 | 87.33 | 84.67 | 87.70 | Last fully connected layer | 90.67 | 84.67 | 85.33 | 78.67 | 88.00 | 89.33 | 90.67 | 84.67 | 79.33 | 81.33 | 90.67 | 85.75 |
|
|