Research map / Label-efficient and robust learning / Domain adaptation and generalization
Tune It the Right Way: Unsupervised Validation of Domain Adaptation via Soft Neighborhood Density
Kuniaki Saito, Donghyun Kim, Piotr Teterwak, Stan Sclaroff, Trevor Darrell, Kate Saenko
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 |
|---|---|---|---|
| select checkpoints with SND instead of Source Risk | unsupervised DA validation · Closed DA Office+OfficeHome | target accuracy (%): 71.1 -> 74.3 | Improved |
| select with SND instead of Entropy | unsupervised DA validation · Closed DA Office+OfficeHome | target accuracy (%): 71.9 -> 74.3 | Improved |
| select with SND instead of DEV | unsupervised DA validation · Closed DA Office+OfficeHome | target accuracy (%): 71.6 -> 74.3 | Improved |
| select with SND instead of Entropy | unsupervised DA validation · Partial DA OfficeHome | target accuracy (%): 60.1 -> 66.0 | Improved |
| select with SND instead of Source Risk | unsupervised DA validation · Partial DA OfficeHome | target accuracy (%): 62.7 -> 66.0 | Improved |
| select with SND instead of Entropy | unsupervised DA validation · VisDA and DomainNet closed DA | target accuracy (%): 64.9 -> 65.8 | No clear effect |
| select with SND instead of Entropy on ADVENT | unsupervised DA validation · GTA5 to Cityscapes | mIoU: 35.6 -> 40.2 | Improved |
| select with SND instead of Entropy on AdaptSeg | unsupervised DA validation · GTA5 to Cityscapes | mIoU: 39.3 -> 39.5 | No clear effect |
| use last-layer softmax output as input feature instead of middle-layer features (ablation) | SND · Office A to D | within-class variance: 0.437 -> 0.163 | Improved |
| add softmax to last-layer output as input feature (ablation) | SND · Office A to D | within-class variance: 0.215 -> 0.163 | Improved |
| add softmax to last-layer output as input feature (ablation) | SND · OfficeHome A to P | within-class variance: 0.301 -> 0.285 | Improved |
| use SND to choose source domain | SND · OfficeHome source domain selection, Target Real | accuracy (%): 74.6 -> 76.0 | Mixed |
Closest papers by meaning
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.91)
- Category Anchor-Guided Unsupervised Domain Adaptation for Semantic Segmentation (NeurIPS 2019 · similarity 0.90)
- Unsupervised Domain Adaptation With Similarity Learning (CVPR 2018 · similarity 0.89)
- Prior Knowledge Guided Unsupervised Domain Adaptation (ECCV 2022 · similarity 0.89)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.89)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 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.
