AI Paper Atlas

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

Meta Pairwise Relationship Distillation for Unsupervised Person Re-Identification

Haoxuanye Ji, Le Wang, Sanping Zhou, Wei Tang, Nanning Zheng, Gang Hua

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Supervised person re-identification · 47 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
add GCN pairwise pseudo-label estimation instead of similarity threshold (SS) (ablation)MPRD · Market-1501Rank-1: 72.8 -> 83.0Improved
add GCN pairwise pseudo-label estimation instead of similarity threshold (SS) (ablation)MPRD · Market-1501mAP: 39.9 -> 51.1Improved
train GCN with unlabeled data in addition to labeled meta data (ablation)MPRD · Market-1501Rank-1: 75.8 -> 83.0Improved
train GCN with unlabeled data in addition to labeled meta data (ablation)MPRD · Market-1501mAP: 43.1 -> 51.1Improved
add CamStyle data augmentation (ablation)MPRD · Market-1501Rank-1: 73.0 -> 83.0Improved
add CamStyle data augmentation (ablation)MPRD · Market-1501mAP: 39.3 -> 51.1Improved
add 5 labeled identities as meta data jump start for GCN (ablation)MPRD · Market-1501Rank-1: 80.9 -> 83.0Improved
add 5 labeled identities as meta data jump start for GCN (ablation)MPRD · DukeMTMC-reIDmAP: 40.1 -> 43.7Improved
give MLCR the same 5 labeled identities as extra supervision (ablation)MLCR · Market-1501Rank-1: 80.1 -> 80.2No clear effect
use KNN instead of similarity threshold for pseudo labels (ablation)pseudo-label generation · Market-1501Rank-1: 72.8 vs 72.7No clear effect
raise likelihood threshold mu beyond about 0.5 (ablation)MPRD · Market-1501Rank-1 and mAPHurt
MPRD versus MLCRunsupervised Re-ID · Market-1501Rank-1: 80.3 -> 83.0Improved

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