AI Paper Atlas

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

Visual correspondence-based explanations improve AI robustness and human-AI team accuracy

Mohammad Reza Taesiri, Giang Nguyen, Anh Totti Nguyen

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 8 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
Use training-set exemplars with kNN instead of the ResNet-50 classifierImageNet classifiers · Adversarial Patchtop-1 accuracy (%): 55.04 -> 59.30Improved
Rerank kNN candidates using five patch correspondences with EMDkNN · ImageNet-Rtop-1 accuracy (%): 36.18 -> 37.75Improved
Rerank kNN candidates using five patch correspondences with CHMkNN · DAmageNettop-1 accuracy (%): 7.59 -> 8.10Improved
Use EMD-Corr rather than kNNImageNet classifiers · ImageNettop-1 accuracy (%): 74.77 -> 74.93Hurt
Use EMD-Corr rather than kNNCUB with iNaturalist features · CUBtop-1 accuracy (%): 85.46 -> 84.98Hurt
Replace automatically selected patches with five human-defined bird keypoints (ablation)CHM-Corr · CUB (iNaturalist features)top-1 accuracy (%): 83.27 -> 81.54Hurt
Increase predefined bird keypoints from five to 10 (ablation)CHM-Corr+ · CUBtop-1 accuracy (%): 81.51 -> 82.34Improved
Increase predefined bird keypoints from five to 15 (ablation)CHM-Corr+ · CUBtop-1 accuracy (%): 81.51 -> 82.27Improved
Use all 49×49 patch pairs instead of five cross-correlation-selected pairs (ablation)EMD-Corr · CUB (iNaturalist ResNet)accuracy (%): 84.98 -> 85.42Improved
Use uniform rather than cross-correlation patch weights (ablation)EMD-Corr · CUB (iNaturalist ResNet)accuracy (%): 84.98 -> 79.72Hurt
Show patch-correspondence boxes in the explanation (ablation)EMD-Corr human-AI explanations · CUB human studyuser accuracy: 64.12 -> 67.64Improved
Show patch-correspondence boxes in the explanation (ablation)CHM-Corr human-AI explanations · CUB human studyuser accuracy: 65.72 -> 69.72Improved

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