AI Paper Atlas

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

Learning to Generate Novel Domains for Domain Generalization

Kaiyang Zhou, Yongxin Yang, Timothy Hospedales, Tao Xiang

ECCV 2020 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Conditional distribution alignment · 436 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 L2A-OT novel-domain generationVanilla · Digits-DG avgaccuracy (%): 73.7 -> 78.1Improved
add L2A-OT novel-domain generationVanilla · PACS avgaccuracy (%): 79.5 -> 82.8Improved
add L2A-OT novel-domain generationVanilla · Office-Home avgaccuracy (%): 64.7 -> 65.6Improved
add L2A-OT novel-domain generationVanilla · Duke->Market1501 re-IDmAP (%): 26.1 -> 30.2Improved
add L2A-OT instead of CrossGrad perturbationCrossGrad · Digits-DG avgaccuracy (%): 75.8 -> 78.1Improved
add CrossGradVanilla · Market1501->Duke re-IDmAP (%): 26.7 -> 27.1No clear effect
generate novel domains instead of source-to-source mapping (StarGAN) (ablation)L2A-OT · Digits-DG avgaccuracy (%): 73.8 -> 78.1Improved
use source-to-source StarGAN data augmentation (ablation)Vanilla · Digits-DG avgaccuracy (%): 73.7 -> 73.8No clear effect
add semantic constraint (cycle-consistency and cross-entropy) (ablation)L2A-OT · Digits-DG avgaccuracy (%): 73.9 -> 78.1Improved
add diversity loss (ablation)L2A-OT · Digits-DG avgaccuracy (%): around 1%Improved
vary number of novel domains Kn (ablation)L2A-OT · Digits-DG avgaccuracy (%)No clear effect
use three source domains instead of two (ablation)L2A-OT · Digits-DG MNIST-Maccuracy (%): 60.9 -> 62.5Improved

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