Table 2 Retrieval results of oracle bone inscription fonts using different methods
From: R-GNN: recurrent graph neural networks for font classification of oracle bone inscriptions
Method | Top-1 accuracy | mAP |
|---|---|---|
ResNet-34 [28] | 0.880 | 0.519 |
GR-RNN(vertical) [16] | 0.960 | 0.656 |
GR-RNN(horizontal) [16] | 0.962 | 0.658 |
FragNet-16 [15] | 0.928 | 0.561 |
FragNet-32 [15] | 0.944 | 0.585 |
FragNet-64 [15] | 0.952 | 0.621 |
Patch [17] | 0.875 | 0.555 |
SA-Net [17] | 0.947 | 0.603 |
MSRF-Net [17] | 0.891 | 0.561 |
HAMVisContexNN [19] | 0.890 | 0.525 |
HAMVisContexNN+WIDNN+Bridge [19] | 0.908 | 0.538 |
WriterINet [20] | 0.950 | 0.619 |
Ours | 0.980 | 0.705 |