Table 2 Ablation study to quantitatively analyze the contribution of each step in the training procedures: ROI segmentation, pre-training for transfer learning, auxiliary co-occurrence loss and ensembling classification models. The F1 scores are shown for each teeth type and all teeth.

From: DeNTNet: Deep Neural Transfer Network for the detection of periodontal bone loss using panoramic dental radiographs

ROI SegmentationPre-trained WeightAuxiliary LossEnsembled NetworkIncisorCaninePremolarMolarAll Teeth
    0.640.670.690.650.66
\(\surd \)   0.710.680.680.670.68
\(\surd \)\(\surd \)  0.730.710.670.690.70
\(\surd \)\(\surd \)\(\surd \) 0.740.720.690.730.72
\(\surd \)\(\surd \)\(\surd \)\(\surd \)0.720.700.750.800.75