1:1 Face Verification on COX Face (V1-S)
77.2TAR (FAR=0.001)Attention-Aware Transformer-Based Aggregation Network
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Attention-Aware Transformer-Based Aggregation NetworkFrame Order=Original, Aggregation Strategy=Transformer-based, Backbone=ResNet-50, Pre-training=VGGFace22026.05 | 77.2 | 92.9 | 99.4 | |
| Attention-Aware Transformer-Based Aggregation NetworkFrame Order=Reversed, Aggregation Strategy=Transformer-based, Backbone=ResNet-50, Pre-training=VGGFace22026.05 | 76.6 | 92.9 | 99.4 | |
| DPR-V2S2026.05 | 75.2 | 83.6 | 94.7 | |
| CNN+MaxPoolAggregation Strategy=Max Pooling, Backbone=ResNet-50, Pre-training=VGGFace22026.05 | 67.2 | 83.5 | 95.2 | |
| CNN+PairwiseAggregation Strategy=Pairwise Comparison, Backbone=ResNet-50, Pre-training=VGGFace22026.05 | 41.7 | 68.6 | 93 | |
| CNN+RandomAggregation Strategy=Random Frame Selection, Backbone=ResNet-50, Pre-training=VGGFace22026.05 | 34.8 | 52.1 | 76.7 |