Research map / Label-efficient and robust learning / Recognition and long-tail
Attentional Feature-Pair Relation Networks for Accurate Face Recognition
Bong-Nam Kang, Yonghyun Kim, Bongjin Jun, Daijin Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add feature-pair bilinear attention without pair selection (ablation) | baseline face encoder · LFW | accuracy (%): 99.60 -> 99.71 | Improved |
| add feature-pair bilinear attention without pair selection (ablation) | baseline face encoder · YTF | accuracy (%): 95.1 -> 96.1 | Improved |
| add top-K feature-pair selection layer (K=442) (ablation) | AFRN · VGGFace2 validation set | accuracy (%): 95.1 -> 97.4 | Improved |
| add top-K feature-pair selection layer (K=442) (ablation) | AFRN · YTF | accuracy (%): 96.1 -> 97.1 | Improved |
| use K=442 instead of using all pairs (K=1,200) (ablation) | AFRN · VGGFace2 validation set | accuracy (%): 2.3% lower | Improved |
| use feature-pair bilinear attention instead of unitary attention (ablation) | AFRN with pair selection · YTF | accuracy (%): 95.3 -> 97.1 | Improved |
| use feature-pair bilinear attention instead of co-attention (ablation) | AFRN with pair selection · LFW | accuracy (%): 99.63 -> 99.85 | Improved |
| add pair selection to AFRN (ablation) | AFRN · IJB-A TAR at FAR=0.001 | TAR: 0.904 -> 0.949 | Improved |
| add pair selection to AFRN (ablation) | AFRN · IJB-B Rank-1 | accuracy: 0.923 -> 0.973 | Improved |
| add pair selection to AFRN (ablation) | AFRN · IJB-B TAR at FAR=0.001 | TAR: 0.933 -> 0.949 | Improved |
| add AFRN without pair selection to baseline encoder (ablation) | model A · IJB-A Rank-1 | accuracy: 0.975 -> 0.981 | Improved |
| add pair selection to AFRN (ablation) | AFRN · CPLFW | accuracy (%): 91.17 -> 93.48 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Pairwise Relational Networks for Face Recognition (ECCV 2018 · similarity 0.89)
- AHAN: Asymmetric Hierarchical Attention Network for Identical Twin Face Verification (AAAI 2026 · similarity 0.85)
- Relation-Aware Global Attention for Person Re-Identification (CVPR 2020 · similarity 0.85)
- Hierarchical Pyramid Diverse Attention Networks for Face Recognition (CVPR 2020 · similarity 0.85)
- CoupleFace: Relation Matters for Face Recognition Distillation (ECCV 2022 · similarity 0.84)
- PairDETR : Joint Detection and Association of Human Bodies and Faces (CVPR 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
