Table 5 Comparison results of different fusion strategies on the Dongba script dataset

From: DFS: Dual-branch forward-looking simulation network for incremental learning of ancient Chinese characters

Method

Accuracy in each session↑

 
 

0

1

2

3

4

5

6

7

8

PD↓

Cutmix

96.251

96.553

96.556

95.421

95.325

94.214

93.109

91.84

90.976

5.275

Random Erasing

96.078

95.734

95.889

94.947

94.955

93.400

92.565

90.965

90.896

5.182

PatchMix

96.804

96.737

96.314

95.368

95.29

94.36

93.152

91.779

91.802

5.002

Augmix

96.11

96.043

95.839

95.211

95.155

93.782

92.913

91.215

90.284

5.826

Mixup

96.639

96.612

96.284

95.425

95.325

94.311

93.50

91.946

91.899

4.74

Back to article page