Table 4 Evaluation metrices.
From: Novel 59-layer dense inception network for robust deepfake identification
Metric | Formula | Focus |
|---|---|---|
Accuracy | (TP + TN) / (TP + TN + FP + FN) | Overall correctness |
Precision | TP / (TP + FP) | Deepfake prediction accuracy |
Recall | TP / (TP + FN) | Deepfake detection ability |
F1-Score | 2TP / (2TP + FP + FN) | Balance between Precision & Recall |
Specificity | TN / (TN + FP) | Real class detection |
False Positive Rate (FPR) | FP / (FP + TN) | False alarms on real data |
ROC-AUC | Area under TPR vs. FPR curve | Overall classification ability |
PR-AUC | Area under Precision vs. Recall curve | Minority class performance |
