R-GNN: recurrent graph neural networks for font classification of oracle bone inscriptions

Heritage Science

Table 1 Classification results of oracle bone inscription fonts using different methods

Method	Top-1	Precision	Recall	F1-score
ResNet-34 [28]	0.804	0.790	0.784	0.787
GR-RNN(vertical) [16]	0.852	0.832	0.831	0.832
GR-RNN(horizontal) [16]	0.857	0.844	0.834	0.839
FragNet-16 [15]	0.833	0.816	0.818	0.817
FragNet-32 [15]	0.845	0.837	0.855	0.845
FragNet-64 [15]	0.847	0.856	0.837	0.846
Patch [17]	0.832	0.829	0.796	0.812
SA-Net [17]	0.860	0.854	0.852	0.853
MSRF-Net [17]	0.836	0.838	0.805	0.821
HAMVisContexNN [19]	0.810	0.805	0.792	0.798
HAMVisContexNN+WIDNN+Bridge [19]	0.822	0.836	0.801	0.818
WriterINet [20]	0.847	0.850	0.847	0.848
Ours	0.882	0.881	0.879	0.880