Research map / Label-efficient and robust learning / Domain adaptation and generalization
Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| divide target data into source-like and target-specific with tailored contrastive goals (DaC) instead of single-scheme contrast (ablation) | SHOT-based adaptation · VisDA-2017 | avg accuracy: 84.4 -> 87.3 | Helped |
| add class-centroid contrastive scheme (Scheme-S) without dividing (ablation) | SHOT · VisDA-2017 | avg accuracy: 82.9 -> 84.1 | Helped |
| add adaptive contrastive loss L_con (ablation) | backbone + L_self · VisDA-2017 | avg accuracy: 83.3 -> 86.5 | Helped |
| add self-training loss L_self (ablation) | backbone · VisDA-2017 | avg accuracy: 59.5 -> 83.3 | Helped |
| add exponential memory-bank MMD (L_EMMD) distribution alignment (ablation) | DaC · VisDA-2017 | avg accuracy: 86.5 -> 87.3 | Helped |
| use exponential MMD instead of linear MMD (ablation) | DaC · VisDA-2017 | avg accuracy: 87.0 -> 87.3 | Helped |
| add local structure positive prototype for target-specific samples (ablation) | DaC · VisDA-2017 | avg accuracy: 86.7 -> 87.3 | Helped |
| add strong augmentation (ablation) | DaC · VisDA-2017 | avg accuracy: 85.6 -> 87.3 | Helped |
| vary confidence threshold tau_c from 0.91 to 0.98 (ablation) | DaC · VisDA-2017 | avg accuracy: 87.06 -> 87.39 | No clear effect |
| add second-stage SHOT++ training (DaC++) | DaC · VisDA-2017 | avg accuracy: 87.3 -> 88.6 | Helped |
Closest papers by meaning
- Contrastive Adaptation Network for Unsupervised Domain Adaptation (CVPR 2019 · similarity 0.91)
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.90)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.90)
- Diffusion-Driven Progressive Target Manipulation for Source-Free Domain Adaptation (NeurIPS 2025 · similarity 0.90)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.90)
- Dynamic Target Distribution Estimation for Source-Free Open-Set Domain Adaptation (AAAI 2025 · similarity 0.89)
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.