Research map / Label-efficient and robust learning / Domain adaptation and generalization
Exploiting Local Feature Patterns for Unsupervised Domain Adaptation
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 |
|---|---|---|---|
| Add local feature alignment to holistic alignment (ablation) | Ours (H) · Office-31 average | accuracy (%): 71.32 -> 72.46 | Helped |
| Add local feature alignment to holistic alignment (ablation) | Ours (H) · Office-home average | accuracy (%): 51.54 -> 53.10 | Helped |
| Add local feature alignment to holistic alignment (ablation) | Ours (H) · Office-31 31-to-25 classes average | accuracy (%): 67.58 -> 69.09 | Helped |
| Use conv5_1 features instead of conv5_3 (ablation) | Proposed method · Office-31 average | accuracy (%): 72.46 -> 56.89 | Hurt |
| Use conv5_2 features instead of conv5_3 (ablation) | Proposed method · Office-31 average | accuracy (%): 72.46 -> 63.67 | Hurt |
| Increase local pattern count from 0 to 32 (ablation) | Proposed method · Office-31 average | accuracy (%): 67.92 -> 72.46 | Helped |
| Increase local pattern count from 32 to 64 (ablation) | Proposed method · Office-31 average | accuracy (%): 72.46 -> 72.59 | Helped |
| Use local feature patterns and joint adaptation rather than holistic-only baselines | Office-31 domain adaptation · Office-31 average | accuracy (%): 67.88 -> 72.46 | Helped |
| Use local feature patterns and joint adaptation rather than holistic-only baselines | Office-home domain adaptation · Office-home average | accuracy (%): 49.78 -> 53.10 | Helped |
| Apply holistic domain adaptation with DANN under source-target label-space mismatch | DANN · Office-31 31-to-25 classes average | accuracy (%): 61.44 -> 58.85 | Hurt |
| Apply holistic domain adaptation with ADDA under source-target label-space mismatch | ADDA · Office-31 31-to-25 classes average | accuracy (%): 61.44 -> 51.10 | Hurt |
| Apply proposed domain adaptation using local patterns | Ours (H+L) · Office-31 31-to-25 classes average | accuracy (%): 61.44 -> 69.09 | Helped |
Closest papers by meaning
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · similarity 0.92)
- Unsupervised Domain Adaptation With Hierarchical Gradient Synchronization (CVPR 2020 · similarity 0.92)
- Joint Domain Alignment and Discriminative Feature Learning for Unsupervised Deep Domain Adaptation (AAAI 2019 · similarity 0.91)
- Unsupervised Domain Adaptation Using Feature-Whitening and Consensus Loss (CVPR 2019 · similarity 0.91)
- Structure-Aware Feature Fusion for Unsupervised Domain Adaptation (AAAI 2020 · similarity 0.91)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · 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.