AI Paper Atlas

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

Comparator Networks

Weidi Xie, Li Shen, Andrew Zisserman

ECCV 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Margin-based face recognition losses · 74 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 Deep Comparator Network (diversity regularizer) on top of SENet50SENet50 · IJB-BTAR at FAR=1E-4: 0.800 -> 0.849Improved
add Deep Comparator Network (keypoints regularizer) on top of ResNet50ResNet50 · IJB-BTAR at FAR=1E-4: 0.784 -> 0.850Improved
add Deep Comparator Network (diversity regularizer) on top of SENet50SENet50 · IJB-CTAR at FAR=1E-4: 0.840 -> 0.885Improved
add Deep Comparator Network (diversity regularizer) on top of ResNet50ResNet50 · IJB-CTAR at FAR=1E-4: 0.825 -> 0.880Improved
ensemble ResNet50 with SENet50single image classification models · IJB-BTAR at FAR=1E-4: 0.800 -> 0.800No clear effect
ensemble ResNet50 with SENet50single image classification models · IJB-CTAR at FAR=1E-4: 0.840 -> 0.841No clear effect
replace diversity regularizer with keypoints regularizer (ablation)SENet50+DCN · IJB-BTAR at FAR=1E-4: 0.849 -> 0.846No clear effect
replace diversity regularizer with keypoints regularizer (ablation)ResNet50+DCN · IJB-CTAR at FAR=1E-4: 0.880 -> 0.867Hurt
replace diversity regularizer with keypoints regularizer (ablation)SENet50+DCN · IJB-CTAR at FAR=1E-1: 0.998 -> 0.998No clear effect
DCN with hard-sample mining vs multicolumn network MN-vcSENet50+DCN(Divs) · IJB-BTAR at FAR=1E-4: 0.831 -> 0.849Improved

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