Table 4.
Fine-tuning and training optimization parameters utilized in the proposed vision-based RD approach.
| Parameter | Value |
|---|---|
| Learning ratio (LR) | 0.00001 |
| Optimization approach | ADAM |
| Regularization approach | L2-regularizer |
| Regularization decay rate | 0.001 |
| Number of epochs | 20 |
| Minimum batch size | 16 |
| Validation frequency | 16 |
| Dropout rate | 0.5 |
| LR schedule parameter | Piecewise |
| LR drop period parameter | 3 |
| LR drop factor parameter | 0.9 |
| Loss function | Categorical cross-entropy |
| Shuffling scenario | Performed every epoch |