Figure 3
From: Generalizability of deep learning models for dental image analysis
(a) The model performance on the Charité and KGMU test sets with different fractions of KGMU images in the training set (markers and error bars denote the mean and standard deviation of the scores over the set of the 5 best models selected from cross-validation, respectively). (b) Model performance for the subsamples of the KGMU dataset with and without root-canal fillings and restorations being present. PPV Positive predictive value.
