AI Paper Atlas

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

Explanation-based Data Augmentation for Image Classification

Sandareka Wickramanayake, Wynne Hsu, Mong-Li Lee

NeurIPS 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 10 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
drop the concept-match term Delta from the utility score (ablation)BRACE with CCNN · CUB Dense-161accuracy (%): 86.8 -> 85.9Hurt
drop the under-representation score beta from the utility score (ablation)BRACE with CCNN · CUB Dense-161accuracy (%): 86.8 -> 86.3Hurt
drop Delta from utility (ablation)BRACE with CCNN · CUB-Families Dense-161accuracy (%): 91.9 -> 91.2Hurt
drop beta from utility (ablation)BRACE with CCNN · CUB-Families Dense-161accuracy (%): 91.9 -> 91.3Hurt
select added samples by utility score instead of randomlyexplanation-based augmentation · CUB-Families Res-101 post-hocaccuracy (%): 88.3 -> 91.2Improved
select samples by classifier confidence instead of utilitysample selection · CUB-Families class accuracy (Hirundinidae)accuracy (%): 41.1 -> 32.9Hurt
augment with BRACE-selected images instead of CutMix/SnapMix mixingCCNN ResNet-34 · CUBaccuracy (%): 80.6 -> 86.1Improved
add Metaset-based repository samplesResNet-34 black-box · Tiny ImageNetaccuracy (%): 76.1 -> 73.1Hurt
augment with BRACEResNet-34 black-box · Tiny ImageNetaccuracy (%): 76.1 -> 78.3Improved
augment with BRACE for out-of-distribution generalizationResNet-34 · ImageNet-V2-Subaccuracy (%): 54.4 -> 70.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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