Research map / Label-efficient and robust learning / Recognition and long-tail
From Face Recognition to Models of Identity: A Bayesian Approach to Learning about Unknown Identities from Unsupervised Data
Daniel Coelho de Castro, Sebastian Nowozin
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 |
|---|---|---|---|
| use the Bayesian identity model instead of NN-distance or one-class SVM for unknown-person detection | OpenFace embeddings on LFW · LFW unknown person detection | AUC: all methods score 95%+ AUC, ours outperforms by a small margin | Improved |
| add the context model (C=3 vs C=1) (ablation) | HDP identity model, identity discovery simulation · Simulated identity discovery (adjusted Rand index) | adjusted Rand index: marginal improvements and reduced variance | Improved |
| process data online (frame by frame) rather than batch or offline (ablation) | HDP identity model Gibbs sampler · Simulated identity discovery (adjusted Rand index) | adjusted Rand index variance: online processing produced much lower variance than batch or offline | Improved |
| use the proposed label model instead of NN or label propagation for acquainted identities | semi-supervised labelling on LFW · LFW semi-supervised labelling (acquaintances) | label prediction accuracy: NN and LP performed nearly identically and slightly better than ours | Hurt |
| increase the number of labelled training images per acquaintance (1 to 15) (ablation) | all labelling methods · LFW semi-supervised labelling | label prediction accuracy: from 1 to 15 labelled images | Improved |
Closest papers by meaning
- Unsupervised and Semi-Supervised Bias Benchmarking in Face Recognition (ECCV 2022 · similarity 0.88)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.87)
- $\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition (NeurIPS 2024 · similarity 0.86)
- Generative Unlearning for Any Identity (CVPR 2024 · similarity 0.85)
- GIF: Generative Inspiration for Face Recognition at Scale (CVPR 2025 · similarity 0.85)
- Person De-reidentification: A Variation-guided Identity Shift Modeling (CVPR 2025 · 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.
