Table 4.
The test accuracies with different partial fine-tuning strategies for VGG-16.
| Model | Fine-tuning targets | Accuracy(%) |
|---|---|---|
| VGG-16 - M1 | All FC layers | 76.05 |
| VGG-16 - M2 | 5th Conv. block + All FC layers | 87.26 |
| VGG-16 - M3 | 4–5th Conv. blocks + All FC layers | 88.51 |
| VGG-16 - M4 | 2–5th Conv. blocks + All FC layers | 93.11 |
| VGG-16 - ALL | All Conv. blocks and FC layers | 97.19 |
In all settings, models are initialized with pre-trained weights from the ImageNet dataset. During the training with the ADAM optimizer, the learning rates are set to 5e-6 and reduced by 0.25 every 15 epochs.