AI Paper Atlas

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

Deep Domain Generalization via Conditional Invariant Adversarial Networks

Ya Li, Xinmei Tian, Mingming Gong, Yajing Liu, Tongliang Liu, Kun Zhang, Dacheng Tao

ECCV 2018 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Optimization-based domain generalization · 634 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 marginal domain-adversarial loss to Deep-All (ablation)DeepA · Rotated MNIST (target 90 deg)accuracy (%): 79.56 -> 82.95Improved
add class-conditional domain loss to marginal-adversarial (DeepC vs DeepD) (ablation)DeepD · Rotated MNIST (target 90 deg)accuracy (%): 82.95 -> 80.08Hurt
add class-conditional domain loss (DeepC vs DeepD) (ablation)DeepD · PACS (target S)accuracy (%): 60.76 -> 64.45Improved
add class-conditional domain loss (DeepC vs DeepD) (ablation)DeepD · VLCS (target C)accuracy (%): 86.58 -> 87.47Improved
add class prior-normalized loss (DeepN vs DeepC) (ablation)DeepC · PACS (target P)accuracy (%): 80.72 -> 77.45Hurt
use class prior-normalized loss instead of class-conditional loss (ablation)DeepC · Rotated MNIST (target 90 deg)accuracy (%): 80.08 -> 83.99Improved
combine class-conditional and prior-normalized losses (CIDDG vs DeepC) (ablation)DeepC · VLCS (target C)accuracy (%): 87.47 -> 88.83Improved
combine both losses (CIDDG vs DeepN) (ablation)DeepN · Rotated MNIST (target 90 deg)accuracy (%): 83.99 -> 84.00No clear effect
combine both losses (CIDDG vs DeepN) (ablation)DeepN · PACS (target P)accuracy (%): 77.45 -> 78.65Improved
CIDDG instead of marginal-matching DeepD (ablation)DeepD · PACS (target P)accuracy (%): 80.39 -> 78.65Hurt
CIDDG instead of marginal-matching DeepD (ablation)DeepD · VLCS (target S)accuracy (%): 60.29 -> 62.10Improved
end-to-end deep fine-tuning instead of shallow methods on pretrained features (DeepA vs MATE)MATE · VLCS (target V)accuracy (%): 51.66 -> 62.71Improved

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