Open Access
Table 1
A summary of the basic parameters in Section 4.1.
| Parameters | Corresponding meanings | Corresponding assignments |
|---|---|---|
| Data | ||
| N | The number of total samples of each dataset | 43 000 |
| Ntest | The count of test samples of each dataset | 4300 |
| M | The number of labeled samples of each dataset | 900 |
| K | The count of challenging samples judged by the initial model | 100 |
| H | The height of a image stamp | 63 pixels |
| W | The width of a image stamp | 63 pixels |
| Model | ||
| Random horizontal flipping | Data augmentation | 0.5 |
| Random vertical flipping | Data augmentation | 0.5 |
| Batch size | The number of samples taken by an iteration | 256 |
| epoch | The number of times for training the entire dataset | 100 |
| τ | Predetermined threshold | 0.95 |
| R | The number of repetitions of the SSLS phase | 3 |
| Learning rate | The step size of the gradient moving toward the optimal Solution of the loss function in each iteration | 2e-4 |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.