AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Domain Generalization with MixStyle

Kaiyang Zhou, Yongxin Yang, Yu Qiao, Tao Xiang

ICLR 2021 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Optimization-based domain generalization · 41 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
mix feature statistics across instances using MixStyleResNet-18 · PACS leave-one-domain-outaverage accuracy: 79.5 -> 83.7Improved
mix feature statistics with random batch shuffling rather than domain labelsResNet-18 · PACS leave-one-domain-outaverage accuracy: 79.5 -> 82.8Improved
mix feature statistics with MixStyleCNN · Digits-DG leave-one-domain-outaverage accuracy: 73.7 -> 76.5Improved
mix feature statistics with MixStyleResNet-18 · Office-Home leave-one-domain-outaverage accuracy: 64.7 -> 65.5Improved
add MixStyleResNet-50 · Market1501→DukemAP: 19.3 -> 23.8Improved
add MixStyleResNet-50 · Duke→Market1501mAP: 20.4 -> 24.7Improved
add MixStyleOSNet · Market1501→DukemAP: 25.9 -> 27.3Improved
add MixStyleOSNet · Duke→Market1501mAP: 24.0 -> 29.0Improved
apply MixStyle to residual blocks 1, 2, and 3 rather than only block 1 (ablation)ResNet-18 · PACS category classificationaccuracy: 80.1 -> 82.8Improved
also apply MixStyle after residual block 4 (ablation)ResNet-18 · PACS category classificationaccuracy: 82.8 -> 75.6Hurt
apply MixStyle at multiple lower-level blocks rather than blocks 1 and 4 (ablation)ResNet-18 · PACS category classificationaccuracy: 76.3 -> 81.7Improved
apply MixStyle after blocks 1 and 2 rather than only block 1 (ablation)ResNet-50 · Cross-dataset person re-IDmAP: 22.6 -> 23.8Improved

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