Research map / Label-efficient and robust learning / Domain adaptation and generalization
Dual Masked Generative Adversarial Transformer for Unsupervised Domain Adaptation
Lei Zhu, Xinxing Xu, Jun Zhou, Qiegen Liu, Rick Siow Mong Goh, Yong Liu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add masked generative adversarial adaptation instead of FADA (ablation) | source-supervised ViT-B · DomainNet clp<->inf | accuracy: 50.5 -> 51.7 | Improved |
| add MGAA on top of masked target domain learning (ablation) | MTDL · DomainNet clp<->inf | accuracy: 54.2 -> 55.3 | Improved |
| reduce both information gap and distribution discrepancy instead of information gap only (ablation) | MTDL + MGAA · DomainNet clp<->inf | accuracy: 54.7 -> 55.3 | Improved |
| use [CLS] token feature reconstruction instead of pixel reconstruction (ablation) | MGAA · DomainNet clp<->inf | accuracy: 47.6 -> 51.7 | Improved |
| use adversarial loss instead of l2 reconstruction loss (ablation) | MGAA · DomainNet clp<->inf | accuracy: 49.7 -> 51.7 | Improved |
| pixel reconstruction with l2 loss added to source training (ablation) | source-supervised ViT-B · DomainNet clp<->inf | accuracy: 45.4 -> 44.1 | Hurt |
| add masked source to target adversarial adaptation branch (dual) (ablation) | Dual-MGAT · DomainNet clp<->inf | accuracy: 51.7 -> 52.9; 55.3 -> 56.0 | Improved |
| Dual-MGAT vs PMTrans | Dual-MGAT · DomainNet (30 tasks) | accuracy: 52.4 -> 58.2 | Improved |
| Dual-MGAT vs MIC(SDAT) | Dual-MGAT · Office-Home | accuracy: 86.2 -> 89.1 | Improved |
| Dual-MGAT vs PMTrans on cataract domain shift | Dual-MGAT · Cataract Slit-Lamp to Camera | AUC: 85.9 -> 94.8 | Improved |
Closest papers by meaning
- Adversarial Discriminative Domain Adaptation (CVPR 2017 · similarity 0.89)
- Unsupervised Pixel-Level Domain Adaptation With Generative Adversarial Networks (CVPR 2017 · similarity 0.89)
- Mind the Discriminability: Asymmetric Adversarial Domain Adaptation (ECCV 2020 · similarity 0.88)
- Bridging Theory and Practice in Source-Free Domain Adaptation via Adversarial Proxy Perturbation (ECCV 2026 · similarity 0.88)
- f-Domain Adversarial Learning: Theory and Algorithms (ICML 2021 · similarity 0.88)
- Domain Adaptation with Asymmetrically-Relaxed Distribution Alignment (ICML 2019 · similarity 0.88)
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.
