AI Paper Atlas

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

Counterfactual Attention Learning for Fine-Grained Visual Categorization and Re-Identification

Yongming Rao, Guangyi Chen, Jiwen Lu, Jie Zhou

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 420 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 counterfactual attention learningmulti-head attention baseline · CUB200-2011top-1 accuracy: 89.3 -> 90.6Improved
add counterfactual attention learningmulti-head attention baseline · Stanford Carstop-1 accuracy: 94.0 -> 95.5Improved
add counterfactual attention learningmulti-head attention baseline · FGVC Aircrafttop-1 accuracy: 93.6 -> 94.2Improved
add counterfactual attention learningResNet50 attention baseline · MSMT17mAP: 50.5 -> 56.2Improved
add counterfactual attention learningResNet50 attention baseline · Market1501mAP: 85.9 -> 87.0Improved
add counterfactual attention learningResNet50 attention baseline · DukeMTMC-ReIDmAP: 74.0 -> 76.4Improved
add counterfactual attention learningResNet50 attention baseline · VehicleID (test 800)Rank-1: 76.7 -> 82.5Improved
add counterfactual attention learningResNet50 attention baseline · VeRi-776mAP: 72.0 -> 74.3Improved
use reversed attention as the counterfactual instead of random (ablation)CAL · CUB200-2011top-1 accuracy: 90.6 -> 89.2Hurt
use uniform attention as the counterfactual instead of random (ablation)CAL · CUB200-2011top-1 accuracy: 90.6 -> 90.2No clear effect
use shuffle attention as the counterfactual instead of random (ablation)CAL · CUB200-2011top-1 accuracy: 90.6 -> 90.4No clear effect
add attention dropout instead of CAL (ablation)attention baseline · CUB200-2011top-1 accuracy: 89.3 -> 88.9Hurt

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