Table 7 Performance comparison on bronze inscription dataset
From: Multi-modal ancient scripts recognition via deep learning with data homogenization and augmentation
Model | Top-1 Acc | F1 Score | AUC | Top-5 Acc | ||||
|---|---|---|---|---|---|---|---|---|
Base | CA | Base | CA | Base | CA | Base | CA | |
AlexNet | 0.631 | 0.763 | 0.610 | 0.778 | 0.962 | 0.991 | 0.842 | 0.942 |
VGG19 | 0.588 | 0.727 | 0.568 | 0.721 | 0.947 | 0.983 | 0.831 | 0.911 |
ResNet50 | 0.615 | 0.803 | 0.615 | 0.797 | 0.965 | 0.994 | 0.850 | 0.960 |
ConvNext | 0.573 | 0.778 | 0.555 | 0.772 | 0.949 | 0.991 | 0.792 | 0.942 |
EfficientNet | 0.623 | 0.744 | 0.613 | 0.739 | 0.962 | 0.983 | 0.850 | 0.921 |
ShuffleNet | 0.531 | 0.754 | 0.514 | 0.746 | 0.923 | 0.987 | 0.762 | 0.931 |
ViT | 0.285 | 0.441 | 0.267 | 0.429 | 0.856 | 0.930 | 0.569 | 0.742 |
SwinTransformer | 0.427 | 0.718 | 0.418 | 0.711 | 0.931 | 0.985 | 0.742 | 0.911 |