Table 8 Baseline models vs proposed method with only data homogenization module (marked as DH only) and CA accuracy comparison on oracle bone inscriptions dataset
From: Multi-modal ancient scripts recognition via deep learning with data homogenization and augmentation
Model | Top-1 Acc | Top-5 Acc | ||||
|---|---|---|---|---|---|---|
Base | DH only | CA | Base | DH only | CA | |
AlexNet | 0.657 | 0.797 | 0.861 | 0.876 | 0.940 | 0.966 |
VGG19 | 0.643 | 0.735 | 0.904 | 0.853 | 0.904 | 0.978 |
ResNet50 | 0.671 | 0.862 | 0.900 | 0.863 | 0.969 | 0.980 |
ConvNext | 0.644 | 0.730 | 0.880 | 0.840 | 0.915 | 0.973 |
EfficientNet | 0.693 | 0.775 | 0.863 | 0.870 | 0.917 | 0.965 |
ShuffleNet | 0.538 | 0.713 | 0.902 | 0.756 | 0.901 | 0.978 |
ViT | 0.288 | 0.386 | 0.774 | 0.519 | 0.691 | 0.901 |
SwinTransformer | 0.421 | 0.562 | 0.858 | 0.662 | 0.806 | 0.973 |
