AI Paper Atlas

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

BackMix: Regularizing Open Set Recognition by Removing Underlying Fore-Background Priors

Yu Wang, Junxian Mu, Hongzhi Huang, Qilong Wang, Pengfei Zhu, Qinghua Hu

TPAMI 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 21 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 BackMixPlain* softmax baseline · CIFAR10 (6 known/4 unknown)AUROC: 67.7 -> 91.3Improved
add BackMixARPL · CIFAR10 (6 known/4 unknown)AUROC: 89.8 -> 91.0Improved
add BackMixARPL · Tiny-ImageNetAUROC: 76.0 -> 76.3No clear effect
add BackMixCSSR · CIFAR10 (6 known/4 unknown)AUROC: 90.7 -> 94.2Improved
replace BackMix with CutmixPlain* softmax · ImageNet30 splitAUIN: 86.4 -> 76.7Hurt
replace BackMix with MixupPlain* softmax · ImageNet30 splitAUIN: 86.4 -> 80.8Hurt
add CutoutPlain* softmax · ImageNet30 splitAUROC: 89.9 -> 90.4No clear effect
use backgrounds from any random image instead of only different-class images (ablation)BackMix · CIFAR10 (6 known/4 unknown)AUROC: 90.9 -> 91.3No clear effect
train on FG+BG* (foreground with random other backgrounds) instead of raw images (ablation)COCO 12-class ResNet18 · COCO known/raw testAUROC: 67.2 -> 69.5Improved
train on FG only instead of raw images (ablation)COCO 12-class ResNet18 · COCO known/raw testAUROC: 67.2 -> 52.7Hurt
train on FG+BG* instead of raw, test with backgrounds replaced (ablation)COCO 12-class ResNet18 · COCO FG+BG*/Raw testAUROC: 59.0 -> 83.6Improved
add outlier exposure with DTD outliers (ablation)Plain WRN-40-4 · CIFAR10 knownAUROC: 87.9 -> 77.4Hurt

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