AI Paper Atlas

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

Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval

Yunquan Zhu, Xinkai Gao, Bo Ke, Ruizhi Qiao, Xing Sun

ICCV 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 14 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
replace ArcFace with MadaCos adaptive loss (ablation)CFCD R50 · ROxford HardmAP: 58.53 -> 64.38Improved
add triplet loss with local descriptor matching after global training (ablation)CFCD R50 · RParis HardmAP: 80.21 -> 81.68Improved
apply triplet loss from the start of training instead of from epoch 50 (ablation)CFCD R50 · ROxford HardmAP: 64.84 -> 58.29Hurt
add hard negative sampling (ablation)CFCD R50 · ROxford HardmAP: 64.84 -> 65.06Improved
add matching constraints to the triplet loss (ablation)CFCD R50 · ROxford HardmAP: 62.37 -> 65.06Improved
use MadaCos instead of ArcFace in the coarse-to-fine framework (ablation)CFCD R50 · ROxford HardmAP: 60.46 -> 65.06Improved
use dynamically tuned scale and margin instead of fixed converged values (ablation)MadaCos · ROxford HardmAP: 50.28 -> 63.36Improved
use MadaCos instead of AdaCos (ablation)global loss · ROxford HardmAP: 56.62 -> 63.36Improved
use a smaller matching threshold tau (ablation)triplet matching · ROxford/RParis MediummAP: tau 30: 82.42 vs tau 70: 81.88 on ROxford MediumMixed
use fewer negatives per triplet (Q) (ablation)triplet loss · ROxford HardmAP: 64.83 (Q=10) -> 66.03 (Q=2)Improved
increase MadaCos anchor rho beyond 0.1 (ablation)MadaCos · ROxford HardmAP: 61.25 (rho 0.1) -> 57.02 (rho 0.4)Hurt
second-stage fine-tuning with longer total training (T=25 vs E=T=20) (ablation)R50-CFCD 2048-d · ROxford HardmAP: 59.99 -> 63.59Improved

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