Table 3
Comparative results of segmentation models.
| Model | DA | DSC | mAP | AUC ROC |
|---|---|---|---|---|
| UNet | no | 0.9270 | 0.9551 | 0.9782 |
| PE-UNet-L | no | 0.9338 | 0.9627 | 0.9819 |
| UNet | yes | 0.9680 | 0.9949 | 0.9960 |
| PE-UNet-L | yes | 0.9746 | 0.9970 | 0.9976 |
Notes. Comparative results of segmentation models with respect to the three classification metrics: DSC which is threshold dependent, and mAP and AUC ROC which are integrated over the threshold range. All results are averaged over 5 folds. The usage of data augmentation virtually increases the dataset size by 16 but breaks the orientation-position relation of filaments.
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.