Table 1 Tooth instance segmentation metrics.The results on the held-out test split of PRs from the OdontoAI platform are averaged for five models trained with 5-fold cross-validation. mAP@IoU=\(x\) denotes mean average precision at intersection over union threshold(s) \(x\)
From: Combining public datasets for automated tooth assessment in panoramic radiographs
mAP@IoU = | 0.5 | 0.75 | 0.5:0.95 |
|---|---|---|---|
Class-agnostic | 0.990 | 0.989 | 0.849 |
FDI | 0.993 | 0.985 | 0.848 |