Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain-Specificity Inducing Transformers for Source-Free Domain Adaptation
Sunandini Sanyal, Ashish Ramayee Asokan, Suvaansh Bhambri, Akshay Kulkarni, Jogendra Nath Kundu, R Venkatesh Babu
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 domain-specificity training via query weights (no DRI) (ablation) | DSiT vendor-side · Office-Home SSDA (4 settings) | average accuracy (%): 74.8 -> 75.6 | Improved |
| add domain-representative inputs (patch-shuffled) to domain classifier training (ablation) | DSiT vendor-side · Office-Home SSDA (4 settings) | average accuracy (%): 75.6 -> 76.6 | Improved |
| add domain-specificity training on client side (ablation) | SHOT · Office-Home SSDA (4 settings) | average accuracy (%): 79.0 -> 80.0 | Improved |
| add DRI to client-side domain-specificity training (ablation) | DSiT client-side · Office-Home SSDA (4 settings) | average accuracy (%): 80.0 -> 81.1 | Improved |
| use DRI as a general augmentation for task classifier training with SHOT-B (ablation) | SHOT-B · Office-Home SSDA (4 settings) | average accuracy (%): 79.1 -> 77.4 | Hurt |
| use DRI as a general augmentation for task classifier training with CDTrans (ablation) | CDTrans · Office-Home SSDA (4 settings) | average accuracy (%): 81.5 -> 76.8 | Hurt |
| apply DSiT to source-free adaptation vs SHOT-B on Office-Home | SHOT-B* · Office-Home SSDA | average accuracy (%): 78.1 -> 80.5 | Improved |
| apply DSiT to source-free adaptation vs SHOT-B on DomainNet | SHOT-B · DomainNet SSDA | average accuracy (%): 38.9 -> 42.1 | Improved |
| apply DSiT to multi-source source-free adaptation | SHOT-B* · Office-Home MSDA | average accuracy (%): 83.7 -> 84.7 | Improved |
| use DSiT instead of non-source-free SSRT-B on Office-Home | SSRT-B · Office-Home SSDA | average accuracy (%): 85.4 -> 80.5 | Hurt |
Closest papers by meaning
- Self-Adversarial Disentangling for Specific Domain Adaptation (TPAMI 2023 · similarity 0.89)
- Balancing Discriminability and Transferability for Source-Free Domain Adaptation (ICML 2022 · similarity 0.89)
- Learning to Balance Specificity and Invariance for In and Out of Domain Generalization (ECCV 2020 · similarity 0.88)
- Heuristic Domain Adaptation (NeurIPS 2020 · similarity 0.87)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.87)
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.86)
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.
