Research map / Graph learning / Graph algorithms and clustering
Learning to Cluster Faces on an Affinity Graph
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 |
|---|---|---|---|
| learn cluster patterns with the proposed graph detection and segmentation framework | face clustering · 580K images, 8,573 identities | F-score: 75.01 -> 85.66 | Helped |
| use clusters from the proposed method as pseudo-labels for unlabeled data | face recognition model · MegaFace | Top-1 identification hit rate: 60.29 -> 78.64 | Helped |
| add GCN-S to refine GCN-D cluster proposals (ablation) | face clustering · 580K images, 8,573 identities | F-score | Helped |
| add a second proposal-generation iteration to combine super-vertices (ablation) | face clustering proposal generation · 580K images, 8,573 identities | F-score: 5% | Helped |
| increase proposal generation to a third iteration (ablation) | face clustering proposal generation · 580K images, 8,573 identities | F-score | Helped |
| use mean pooling instead of max pooling (ablation) | GCN-D · 580K images, 8,573 identities | F-score: 83.61 -> 76.97 | Hurt |
| use sum pooling instead of max pooling (ablation) | GCN-D · 580K images, 8,573 identities | F-score: 83.61 -> 53.75 | Hurt |
| remove vertex features (ablation) | GCN-D · 580K images, 8,573 identities | F-score: 83.61 -> 73.06 | Hurt |
| widen the GCN-D channels to 256, 64 (ablation) | GCN-D · 580K images, 8,573 identities | F-score: 83.61 -> 84.16 | Helped |
| use a deeper GCN-D with 256, 128, 64 channels (ablation) | GCN-D · 580K images, 8,573 identities | F-score: 83.61 -> 77.95 | Hurt |
| apply GCN-S to remove outliers from existing clustering outputs (ablation) | K-means, HAC, and CDP clustering results · face clustering | F-score: 2%-5% | Helped |
| use de-overlapping instead of NMS (ablation) | post-processing of face clusters · face clustering | clustering performance | Helped |
Closest papers by meaning
- Learning to Cluster Faces via Confidence and Connectivity Estimation (CVPR 2020 · similarity 0.93)
- Linkage Based Face Clustering via Graph Convolution Network (CVPR 2019 · similarity 0.92)
- Structure-Aware Face Clustering on a Large-Scale Graph With 107 Nodes (CVPR 2021 · similarity 0.91)
- FC-DGCN: Deep Graph Convolutional Network for Face Clustering and Recognition (NeurIPS 2026 · similarity 0.91)
- Density-Aware Feature Embedding for Face Clustering (CVPR 2020 · similarity 0.88)
- Image Clustering via the Principle of Rate Reduction in the Age of Pretrained Models (ICLR 2024 · similarity 0.85)
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.