Research map / Label-efficient and robust learning / Domain adaptation and generalization
De-Confusing Pseudo-Labels in Source-Free Domain Adaptation
Idit Diamant, Amir Rosenfeld, Idan Achituve, Jacob Goldberger, Arnon Netzer
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 prior loss on transition matrix (domain adaptation regularization) to DCPL (ablation) | SHOT w/ DCPL · VisDA | average accuracy: 85.9 -> 86.7 | Improved |
| add prior loss on transition matrix to DCPL (ablation) | SHOT w/ DCPL · DomainNet-126 | average accuracy: 69.3 -> 69.6 | Improved |
| add prior loss on transition matrix to DCPL (ablation) | SHOT w/ DCPL · OfficeHome | average accuracy: 77.4 -> 77.5 | Improved |
| use a stronger pseudo-labeling network (SwinB-22K-1K instead of SwinB-1K) (ablation) | SHOT w/ DCPL · OfficeHome | average accuracy: 77.5 -> 83.8 | Improved |
| use a stronger pseudo-labeling network (SwinB-22K-1K instead of SwinB-1K) (ablation) | SHOT w/ DCPL · DomainNet-126 | average accuracy: 69.6 -> 73.7 | Improved |
| use a stronger pseudo-labeling network (SwinB-22K-1K instead of SwinB-1K) (ablation) | SHOT++ w/ DCPL · OfficeHome | average accuracy: 77.8 -> 83.8 | Improved |
| learn the noise transition matrix instead of fixing it to identity (ablation) | SHOT w/ DCPL · DomainNet-126 (SwinB-1K) | average accuracy: 65.3 -> 69.6 | Improved |
| learn the noise transition matrix instead of fixing it to identity (ablation) | SHOT w/ DCPL · DomainNet-126 (SwinB-22K-1K) | average accuracy: 71.5 -> 73.7 | Improved |
| use the true (oracle) transition matrix instead of the learned one (ablation) | SHOT w/ DCPL · DomainNet-126 (SwinB-1K) | average accuracy: 69.6 -> 72.5 | Improved |
| replace Co-learning cross-entropy with DCPL cross-entropy | SHOT · DomainNet-126 | average accuracy: 67.6 -> 69.6 | Improved |
| replace Co-learning with DCPL | SHOT++ · OfficeHome | average accuracy: 76.1 -> 77.8 | Improved |
| replace Co-learning with DCPL | SHOT++ · DomainNet-126 | average accuracy: 69.3 -> 70.5 | Improved |
Closest papers by meaning
- When Source-Free Domain Adaptation Meets Learning with Noisy Labels (ICLR 2023 · similarity 0.93)
- Label-Noise Robust Domain Adaptation (ICML 2020 · similarity 0.93)
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.90)
- Proxy Denoising for Source-Free Domain Adaptation (ICLR 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)
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.
