Research map / Architectures and efficiency / Attention and transformers
Cluster and Aggregate: Face Recognition with Large Probe Set
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace naive average fusion with Cluster and Aggregate fusion | IResNet101+ArcFace · IJB-B | TAR@FAR=1e-4: 94.30 -> 95.53 | Helped |
| Replace naive average fusion with Cluster and Aggregate fusion | IResNet101+ArcFace · IJB-C | TAR@FAR=1e-4: 95.78 -> 97.15 | Helped |
| Replace naive average fusion with Cluster and Aggregate fusion | ArcFace · IJB-S Surveillance-to-Surveillance | Rank-1: 32.13 -> 36.51 | Helped |
| Add style-based keys instead of clustering directly from identity features (ablation) | CAFace · IJB-B | TAR@FAR=1e-3: 96.11 -> 96.81 | Helped |
| Add feature-norm embedding to style input (ablation) | CAFace with style-only input · IJB-B | TAR@FAR=1e-3: 96.81 -> 96.91 | Helped |
| Add feature-norm embedding to style input (ablation) | CAFace with style-only input · IJB-S | AVG: 57.42 -> 57.55 | Helped |
| Add set permutation consistency loss L_p (ablation) | CAFace · IJB-B | TAR@FAR=1e-3: 96.86 -> 96.91 | No clear effect |
| Add set permutation consistency loss L_p (ablation) | CAFace · IJB-S | AVG: 57.36 -> 57.55 | Helped |
| Increase cluster centers from 1 to 4 (ablation) | CAFace · IJB-S | AVG: 53.87 -> 57.55 | Helped |
| Increase cluster centers from 2 to 4 (ablation) | CAFace · IJB-S | AVG: 57.11 -> 57.55 | Helped |
| Increase cluster centers from 4 to 8 (ablation) | CAFace · IJB-B | TAR@FAR=1e-4: 95.53 -> 95.61 | Helped |
| Increase cluster centers from 4 to 8 (ablation) | CAFace · IJB-S | AVG: 57.55 -> 57.33 | Hurt |
Closest papers by meaning
- Learning to Cluster Faces on an Affinity Graph (CVPR 2019 · similarity 0.87)
- ProxyFusion: Face Feature Aggregation Through Sparse Experts (NeurIPS 2024 · similarity 0.87)
- FaceCoresetNet: Differentiable Coresets for Face Set Recognition (AAAI 2024 · similarity 0.86)
- Neural Aggregation Network for Video Face Recognition (CVPR 2017 · similarity 0.86)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.84)
- RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition (AAAI 2025 · 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.