Table 1 Comparative result between regression models and segmentation models with our proposed module
From: Detecting oracle bone inscriptions via pseudo-category labels
| Â | Precision | Recall | F1-Score | ||||||
|---|---|---|---|---|---|---|---|---|---|
Train | Val | Test | Train | Val | Test | Train | Val | Test | |
RetinaNet [41] | 60.5 | 57.3 | 56.1 | 62.3 | 56.4 | 55.4 | 61.4 | 56.9 | 55.4 |
YOLO [44] | 47.6 | 44.1 | 46.3 | 47.2 | 45.6 | 46.9 | 47.4 | 44.8 | 46.6 |
YOLO(pre-train) | 85.4 | 73.2 | 73.7 | 82.4 | 72.8 | 74.1 | 83.8 | 72.7 | 73.9 |
Hourglass(ours) | 88.9 | 71.3 | 70.9 | 85.5 | 68.6 | 68.4 | 87.1 | 69.9 | 69.6 |
U\(^2\)Net(ours) | 92.7 | 73.8 | 74.1 | 89.4 | 72.1 | 72.5 | 91.0 | 72.9 | 73.4 |
UNet(ours) | 91.6 | 71.9 | 71.5 | 88.6 | 70.3 | 69.8 | 90.1 | 71.1 | 70.6 |