AI Paper Atlas

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

Towards Effective Instance Discrimination Contrastive Loss for Unsupervised Domain Adaptation

Yixin Zhang, Zilei Wang, Junjie Li, Jiafan Zhuang, Zihan Lin

ICCV 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Conditional distribution alignment · 25 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
use low-confidence samples only for IDCo loss instead of all samples (ablation)GVB+FixMatch with IDCo · Office-Home UDAaccuracy: 73.0 -> 73.3Improved
use high-confidence samples for IDCo loss (ablation)GVB+FixMatch with IDCo · Office-Home UDAaccuracy: 72.7 -> 72.3Hurt
add class relationship enhanced features (CRF) to low-confidence IDCo (ablation)EIDCo · Office-Home UDAaccuracy: 73.3 -> 74.4Improved
add CRF to high-confidence samples (ablation)IDCo loss · Office-Home UDAaccuracy: 72.3 -> 72.4No clear effect
add plain cross-domain mixup (ablation)EIDCo · Office-Home UDAaccuracy: 74.4 -> 74.6Improved
use target-dominated mixup rather than plain mixup (ablation)EIDCo · Office-Home UDAaccuracy: 74.6 -> 75.1Improved
use CRF similarity instead of original features for EIDCo (ablation)EIDCo · Office-Home UDAaccuracy: 73.6 -> 75.1Improved
use CRF similarity instead of original features for EIDCo (ablation)EIDCo · VisDA-2017 UDAaccuracy: 80.1 -> 82.0Improved
add EIDCo on top of FixMatch with strong ViT baselineBaseline-B in SSRT + FixMatch · Office-Home UDAaccuracy: 83.07 -> 86.38Improved
add EIDCo alone instead of FixMatch to baselineBaseline-B in SSRT · VisDA-2017 UDAclass-wise accuracy: FixMatch 86.61 vs EIDCo 86.37Hurt
add EIDCo to weaker baselines (Source only, DANN, CDAN)weaker UDA baselines · Office-Home UDAaccuracy: 46.1 -> 52.6; 57.6 -> 60.3; 65.8 -> 67.6Improved
apply IDCo loss directly in feature space (even with projection head)UDA baselines · UDA benchmarksaccuracy: < 0.5%No 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