Table 4 Evaluation results of pre-trained Vision Transformers (ViTs) and their ensembles using weighted and soft voting.

ViT-base³⁹

0.80

0.81

0.80

0.77

0.46

0.8 ms

ViT-large³⁹

0.87

0.87

0.87

0.87

0.62

15.5 ms

Swin-tiny⁴¹

0.85

0.85

0.85

0.84

0.65

4.7 ms

Swin-base⁴¹

0.90

0.90

0.90

0.89

0.77

3.1 ms

DeiT-small⁴⁰

0.83

0.82

0.83

0.82

0.60

10.7 ms

DeiT-base⁴⁰

0.87

0.87

0.87

0.87

0.64

4.7 ms

Vision transformer ensemble

Weighted voting ensemble

0.91

0.91

0.91

0.90

0.76

31.56 ms

Soft voting ensemble

0.93

0.94

0.93

0.93

0.77

7.0 ms