Table 2 The performance of the proposed method with no augmentation (5-fold cross-validation results (.

From: Hybrid framework for image forgery detection and robustness against adversarial attacks using vision transformer and SVM

Dataset

Precision

(mean ± std)

Recall

(mean ± std)

F1-score

(mean ± std)

Accuracy

(mean ± std)

CASIA1

97.12 ± 0.42

96.25 ± 0.2

96.68 ± 0.44

97.85 ± 0.61

CASIA2

96.73 ± 0.80

96.08 ± 0.71

96.4 ± 0.34

97.27 ± 0.27

MICC-F220

97.22 ± 0.41

97.12 ± 0.3

97.16 ± 0.36

97.98 ± 0.34

MICC-F600

96.59 ± 0.52

96.23 ± 0.51

96.4 ± 0.52

97.58 ± 0.54

MICC-F2000

97.09 ± 0.30

96.68 ± 0.20

96.88 ± 0.43

97.85 ± 0.47

Merged Dataset

98.27 ± 0.24

97.37 ± 0.02

97.81 ± 0.23

98.31 ± 0.22

Back to article page