AI Paper Atlas

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

Attributable Visual Similarity Learning

Borui Zhang, Wenzhao Zheng, Jie Zhou, Jiwen Lu

CVPR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 15 citations (OpenAlex)

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 AVSL to ProxyAnchorProxyAnchor · CUB-200-2011Recall@1: 69.7 -> 71.9Improved
Apply AVSL to MarginMargin · CUB-200-2011Recall@1: 65.6 -> 68.8Improved
Apply AVSL to ProxyAnchorProxyAnchor · Cars196Recall@1: 87.7 -> 91.5Improved
Apply AVSL to MarginMargin · Cars196Recall@1: 78.2 -> 81.1Improved
Apply AVSL to ProxyAnchorProxyAnchor · Stanford Online ProductsRecall@1: 78.4 -> 79.6Improved
Apply AVSL to MarginMargin · Stanford Online ProductsRecall@1: 72.4 -> 76.8Improved
Add multi-layer embedding losses (ablation)ProxyAnchor · Cars196Recall@1: 87.7 -> 89.7Improved
Add reliability-based node selection to multi-layer losses (ablation)ProxyAnchor + M · Cars196Recall@1: 89.7 -> 89.9Improved
Add graph relations and hierarchy-consistent AVSL inference (ablation)ProxyAnchor + M & R · Cars196Recall@1: 89.9 -> 91.5Improved
Concatenate layer embeddings instead of using AVSL inference (ablation)ProxyAnchor · Cars196Recall@1: 90.6 -> 91.5Hurt
Remove momentum updating for edge construction (ablation)ProxyAnchor-AVSL · Cars196Recall@1: 91.0 -> 91.5Hurt
Use a larger edge momentum factor, γ=0.95 rather than γ=0.50 (ablation)ProxyAnchor-AVSL · Cars196Recall@1: 91.5 -> 91.6Improved

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