Table 3 Comparison between the proposed method and state-of-the-art methods on CASIA-SURF (Unit: %).
From: Multimodal feature enhancement via dynamically-aware heterogeneous network for face anti-spoofing
Model | APCER | BPCER | ACER | TPR@FPR=10E-2 | TPR@FPR=10E-3 | TPR@FPR=10E-4 |
|---|---|---|---|---|---|---|
Halfway fusion11 | 5.60 | 3.80 | 4.70 | 89.1 | 33.6 | 17.8 |
SE fusion11 | 3.80 | 1.00 | 2.40 | 96.7 | 81.8 | 56.8 |
PipeNet28 | 2.08 | 2.45 | 2.26 | 95.90 | 82.10 | 56.7 |
MA-Net29 | 2.40 | 1.70 | 2.00 | 96.00 | 82.60 | 58.1 |
MFF-CNN30 | 2.84 | 4.12 | 3.48 | – | – | – |
Conv-MLP31 | 1.50 | 1.80 | 1.60 | – | – | – |
MF²ShrT32 | 1.60 | 1.20 | 1.40 | – | – | – |
MFViT and MRF33 | 1.50 | 1.70 | 1.60 | – | – | – |
DACA-CNN34 | 2.77 | 3.13 | 2.95 | – | – | |
ECA-ICD35 | 5.57 | 0.65 | 3.11 | – | – | – |
Ours | 1.31 | 0.72 | 1.01 | 99.19 | 95.35 | 87.32 |
