Table 4 Object detection metrics of the treatment detection task on the test quadrant dataset.
From: Dental enumeration and multiple treatment detection on panoramic X-rays using deep learning
Data type | Loss type | Settings | Area | # Detections | \(AP_{0.50:0.95}\) | \(AR_{0.50:0.95}\) | \(AP_{0.50}\) |
|---|---|---|---|---|---|---|---|
Normal quadrants | Weighted | Flip distortion | All | 100 | 32.6 | 46.6 | 54.6 |
Medium | 1000 | 51.7 | 51.7 | – | |||
Large | 1000 | 32.5 | 46.5 | – | |||
Rotated quadrants | Weighted | Distortion | All | 100 | 37.2 | 52.4 | 55.9 |
Medium | 1000 | 20.0 | 20.0 | – | |||
Large | 1000 | 37.3 | 52.7 | – | |||
Normal quadrants | Weighted class weight | Distortion | All | 100 | 32.5 | 45.8 | 54.3 |
Medium | 1000 | 10.1 | 10.0 | – | |||
Large | 1000 | 32.6 | 45.9 | – | |||
Rotated quadrants | Weighted class weight | Distortion | All | 100 | 36.6 | 50.5 | 58.6 |
Medium | 1000 | 30.0 | 30.0 | – | |||
Large | 1000 | 36.7 | 50.7 | – | |||
Normal quadrants | Weighted | Distortion augmented data* | All | 100 | 35.7 | 50.2 | 57.3 |
Medium | 1000 | 48.4 | 48.3 | – | |||
Large | 1000 | 35.5 | 50.2 | – | |||
Normal quadrants | Weighted | Distortion negative sampling** | All | 100 | 37.7 | 52.1 | 59.0 |
Medium | 1000 | 48.4 | 48.3 | – | |||
Large | 1000 | 37.5 | 52.1 | – |
