Table 22 Performance comparison for in-the-wild and cross-distribution scenarios.
From: PatternFusion: a hybrid model for pattern recognition in time-series data using ensemble learning
Model configuration | In-the-wild dataset | Accuracy (%) | F1-score (%) | AUC (%) | EER (%) |
|---|---|---|---|---|---|
PatternFusion (baseline) | FaceForensics++ | 87.3 | 85.2 | 89.1 | 4.3 |
LFW | 88.9 | 86.7 | 90.5 | 3.9 | |
PatternFusion + fine-tuning | FaceForensics++ | 91.8 | 89.4 | 92.7 | 3.2 |
LFW | 93.2 | 91.1 | 94.1 | 2.8 | |
PatternFusion + domain-adversarial training | FaceForensics++ | 93.5 | 91.6 | 94.3 | 2.5 |
LFW | 94.7 | 93.4 | 95.8 | 2.2 |
