AI Paper Atlas

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

Learning To Diversify for Single Domain Generalization

Zijian Wang, Yadan Luo, Ruihong Qiu, Zi Huang, Mahsa Baktashmotlagh

ICCV 2021 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Foundation-model domain generalization · 314 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 style-complement module (generated images) (ablation)Ours · PACS (train photo, single DG)accuracy (%) Avg.: 45.39 -> 55.24Improved
add style modification (ablation)Ours · PACS (train photo, single DG)accuracy (%) Avg.: 53.54 -> 55.24Improved
add mutual information minimization (ablation)Ours · PACS (train photo, single DG)accuracy (%) Avg.: 53.63 -> 55.24Improved
add mutual information maximization (ablation)Ours · PACS (train photo, single DG)accuracy (%) Avg.: 51.13 -> 55.24Improved
increase number of transformations k up to about 5 (ablation)Ours · PACS (train photo, single DG)accuracy (%)Improved
increase transformations k beyond 5 (ablation)Ours · PACS (train photo, single DG)accuracy (%)No clear effect
vary loss weights alpha1, alpha2 (ablation)Ours · PACS (train photo, single DG)accuracy (%)No clear effect
replace adversarial augmentation (ME-ADA) with style-complement generationME-ADA · Digits (train MNIST)accuracy (%) Avg.: 59.32 -> 74.46Improved
replace ME-ADA with L2D on CIFAR-10-C NoiseME-ADA · CIFAR-10-C (severity 5)accuracy (%) Noise: 66.47 -> 73.29Improved
L2D vs ME-ADA on USPS (similar stroke style to MNIST)ME-ADA · Digits (train MNIST)accuracy (%) USPS: 81.04 -> 83.97No clear effect

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