AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

Test-Time Style Shifting: Handling Arbitrary Styles in Domain Generalization

Jungwuk Park, Dong-Jun Han, Soyeong Kim, Jaekyun Moon

ICML 2023 · Label-efficient and robust learning › Test-time adaptation · approach: Realistic and stable test-time adaptation · 0 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 test-time style shifting to MixStyleMixStyle · PACS originalaverage accuracy: 82.98 -> 84.28Improved
Add style balancing and test-time style shifting to MixStyleMixStyle · PACS cross-domain data imbalanceaverage accuracy: 75.65 -> 80.69Improved
Add style balancing to DSUDSU · PACS cross-domain class imbalanceaverage accuracy: 40.28 -> 51.75Improved
Add test-time style shifting to DSU without style balancingDSU · PACS cross-domain class imbalanceaverage accuracy: 40.28 -> 39.47Hurt
Add style balancing and test-time style shifting to DSUDSU · PACS cross-domain class imbalanceaverage accuracy: 40.28 -> 52.22Improved
Add style balancing and test-time style shifting to MLDGMLDG · PACS cross-domain class imbalanceaverage accuracy: 40.26 -> 53.54Improved
Add style balancing and test-time style shifting to CDANNCDANN · PACS cross-domain class imbalanceaverage accuracy: 25.61 -> 46.22Improved
Select style-balancing samples by similar style statistics instead of randomly (ablation)Style balancing + MixStyle · PACS cross-domain class imbalanceaverage accuracy: 49.51 -> 50.32Improved
Select style-balancing samples by similar style statistics instead of randomly (ablation)TSB + MixStyle · PACS cross-domain class imbalanceaverage accuracy: 49.32 -> 50.83Improved
Shift all test samples, including those with small style gaps (ablation)TSB + MixStyle · PACS originalaverage accuracy: 84.63 -> 84.68No clear effect
Shift to the nearest sample instead of the nearest source-domain center (ablation)TSB + MixStyle · PACS originalaverage accuracy: 84.63 -> 84.14Hurt
Apply style shifting after the 4th residual block rather than the 2nd (ablation)TSB + MixStyle · PACS originalaverage accuracy: 84.63 -> 20.21Hurt

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