Table 2.
Training accuracy for different fine-tuning configurations. The network is first trained on Dataset-1 and Dataset-2 is used to fine-tune the selected layers. The 7-layer configuration with the minimum number of trainable parameters is selected.
Configuration | Dice Coefficient |
---|---|
19-layers | 0.875 |
13-layers | 0.873 |
7-layers | 0.863 |