Table 1 Comparative experiment results (%) from different network structures
From: AttGraph disentangling confusable ancient Chinese characters via component-correlation synergy
Method | Top-1 | Top-2 | Top-3 | Top-4 | Top-5 |
|---|---|---|---|---|---|
HUNet_4833 | 36.13 | 46.96 | 52.87 | 56.19 | 59.02 |
ResNet | 64.94 | 73.64 | 77.49 | 79.98 | 81.69 |
Inception_v4 | 66.94 | 75.30 | 78.90 | 81.06 | 82.61 |
MME16 | 73.23 | 81.55 | 84.94 | 86.76 | 87.91 |
CA-CF17 | 76.31 | 83.25 | 86.16 | 87.72 | 88.73 |
TinyViT28 | 77.07 ± 0.12 | 84.25 ± 0.51 | 87.00 ± 0.59 | 88.47 ± 0.23 | 89.58 ± 0.17 |
AttGraph (ours) | 78.33 ± 0.13 | 85.48 ± 0.16 | 88.13 ± 0.41 | 89.67 ± 0.72 | 90.53 ± 0.12 |
