AI Paper Atlas

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

Coded Residual Transform for Generalizable Deep Metric Learning

SHICHAO KAN, Yixiong Liang, Min Li, Yigang Cen, Jianxin Wang, Zhihai He

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 3 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
apply CRT (ablation)MS metric learning with MiT-B1 · CUB-200-2011Recall@1: 73.97 -> 75.95Improved
add CRT-Consistency (ablation)CRT metric learning with MiT-B1 · CUB-200-2011Recall@1: 72.25 -> 75.95Improved
add multi-perspective CRT feature transformation (ablation)CRT metric learning with MiT-B1 · CUB-200-2011Recall@1: 74.34 -> 75.95Improved
use different embedding networks for each prototype's CRT feature instead of the same embedding network (ablation)CRT metric learning with MiT-B1 · CUB-200-2011Recall@1Improved
vary first-branch prototype count from 1 to 4, 16, or 64 (ablation)CRT metric learning with MiT-B1 · CUB-200-2011Recall@1: 75.76; 75.57; 75.44; 75.96Mixed
vary second-branch prototype count from 1 to 4, 16, 64, or 100 (ablation)CRT metric learning with MiT-B1 · CUB-200-2011Recall@1: 74.26; 74.54; 74.63; 75.95; 75.15Mixed
share weights between the two embedding branches (ablation)CRT metric learning with MiT-B1 · CUB-200-2011Recall@1: 75.96 vs 75.95No clear effect
replace Hyp-DeiT with CRTdeep metric learning · CUB-200-2011Recall@1: 74.7 -> 78.98Improved
replace Group Loss++ with CRTdeep metric learning · Cars-196Recall@1: 90.4 -> 91.16Improved
replace XBM+RTT with CRTdeep metric learning · SOPRecall@1: 84.5 -> 83.41Hurt
replace IBC with CRTdeep metric learning · In-ShopRecall@1: 92.8 -> 94.48Improved

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