AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

CLIP-Cluster: CLIP-Guided Attribute Hallucination for Face Clustering

Shuai Shen, Wanhua Li, Xiaobing Wang, Dafeng Zhang, Zhezhu Jin, Jie Zhou, Jiwen Lu

ICCV 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Margin-based face recognition losses · 17 citations (OpenAlex)

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 changedWhereMeasuredEffect
use neighbor-aware cross-attention fusion instead of averaging hallucinated features (ablation)CLIP-Cluster proxy generator · MS1M Part1Pairwise F-score: 93.76 -> 94.22Improved
use neighbor-aware cross-attention instead of self-attention (ablation)CLIP-Cluster proxy generator · MS1M Part1Pairwise F-score: 94.04 -> 94.22Improved
add Infomap graph refinement instead of naive pruning (ablation)CLIP-Cluster inference · MS1M Part1Pairwise F-score: 88.86 -> 94.22Improved
add attribute hallucination and proxy generator on top of Infomap (ablation)Infomap clustering · MS1M Part1Pairwise F-score: 93.98 -> 94.22Improved
use CLIP-Cluster instead of ADA-Nets on larger test setsface clustering · MS1M 5.21M unlabeledPairwise F-score: 83.99 -> 87.99Improved
use CLIP-Cluster pseudo-labels for face recognition trainingface recognition on MegaFace · MegaFace rank-1, 1M distractorsrank-1 accuracy (%): 58.2% -> 80.5%Improved
use CLIP-Cluster instead of STAR-FC++face clustering · MS1M Part1inference time: 312s -> 280sImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy