Research map / Label-efficient and robust learning / Recognition and long-tail
GroupFace: Learning Latent Groups and Constructing Group-Based Representations for Face Recognition
Check 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 |
|---|---|---|---|
| Add group-aware representations to ArcFace | ArcFace with ResNet-100 · IJB-B | TAR@FAR=1e-6: 0.3828 -> 0.4678 | Helped |
| Add group-aware representations to ArcFace | ArcFace with ResNet-100 · IJB-C | TAR@FAR=1e-6: 0.8625 -> 0.9053 | Helped |
| Add group-aware representations to ArcFace | ArcFace with ResNet-100 · AgeDB-30 | verification accuracy (%): 98.28 -> 98.28 | No clear effect |
| Use 32 groups rather than the baseline system (ablation) | GroupFace · IJB-B | TAR@FAR=1e-6: 0.3828 -> 0.4678 | Helped |
| Add self-grouping loss (ablation) | GroupFace · IJB-B | TAR@FAR=1e-6: 0.4468 -> 0.4678 | Helped |
| Use self-distributed rather than naive group labels (ablation) | GroupFace · IJB-B | TAR@FAR=1e-6: 0.4535 -> 0.4678 | Helped |
| Use soft rather than hard group ensemble (ablation) | GroupFace · IJB-B | TAR@FAR=1e-6: 0.4439 -> 0.4678 | Helped |
| Use soft rather than hard group ensemble (ablation) | GroupFace · IJB-B | GFLOPS: 24.4G -> 24.5G | Hurt |
| Concatenate rather than aggregate representations (ablation) | GroupFace · IJB-B | TAR@FAR=1e-6: 0.4678 -> 0.4745 | Helped |
| Concatenate rather than aggregate representations (ablation) | GroupFace · IJB-B | TAR@FAR=1e-5: 0.9115 -> 0.8999 | Hurt |
| Add group-aware similarity at evaluation (ablation) | GroupFace · IJB-B | TAR@FAR=1e-6: 0.4678 -> 0.5212 | Helped |
| Add GroupFace to a ResNet-34 backbone (ablation) | ResNet-34 · IJB-B | TAR@FAR=1e-6: 0.3962 -> 0.4361 | Helped |
Closest papers by meaning
- Grouped Knowledge Distillation for Deep Face Recognition (AAAI 2023 · similarity 0.88)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.86)
- Perceptual Group Tokenizer: Building Perception with Iterative Grouping (ICLR 2024 · similarity 0.85)
- ID-Patch: Robust ID Association for Group Photo Personalization (CVPR 2025 · similarity 0.85)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.85)
- Masked Face Recognition with Generative-to-Discriminative Representations (ICML 2024 · similarity 0.84)
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.