Research map / Label-efficient and robust learning / Domain adaptation and generalization
Cross-Domain Ensemble Distillation for Domain Generalization
Kyungmoon Lee, Sungyeon Kim, Suha Kwak
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 XDED to vanilla ResNet-18 (ablation) | ResNet-18 · PACS Art (multi-source) | accuracy: 77.0 -> 83.3 | Improved |
| add UniStyle to vanilla ResNet-18 (ablation) | ResNet-18 · PACS Art (multi-source) | accuracy: 77.0 -> 81.2 | Improved |
| add UniStyle on top of XDED (ablation) | XDED · PACS Art (multi-source) | accuracy: 83.3 -> 85.6 | Improved |
| add XDED to vanilla ResNet-18 (ablation) | ResNet-18 · OfficeHome Clipart (multi-source) | accuracy: 49.4 -> 55.2 | Improved |
| add UniStyle to vanilla ResNet-18 (ablation) | ResNet-18 · OfficeHome Clipart (multi-source) | accuracy: 49.4 -> 50.4 | Improved |
| add UniStyle on top of XDED (ablation) | XDED · OfficeHome Clipart (multi-source) | accuracy: 55.2 -> 57.1 | Improved |
| add UniStyle to vanilla ResNet-50 re-ID (ablation) | ResNet-50 · Market1501 to DukeMTMC-reID | mAP: 19.3 -> 26.2 | Improved |
| add XDED to vanilla ResNet-50 re-ID (ablation) | ResNet-50 · Market1501 to DukeMTMC-reID | mAP: 19.3 -> 24.2 | Improved |
| add XDED + UniStyle in place of one-hot-only domain generalization baselines (MixStyle) | ResNet-18 · PACS Cartoon entropy | accuracy: 75.9 -> 81.7 | Improved |
| add XDED + UniStyle versus ResNet-18 under FGSM | ResNet-18 · PACS Photo under FGSM attack (multi-source) | accuracy: 39.6 -> 55.4 | Improved |
| add XDED + UniStyle versus ResNet-18 under PGD | ResNet-18 · PACS Photo under PGD attack (multi-source) | accuracy: 16.3 -> 30.4 | Improved |
| add XDED + UniStyle versus Mixup on corruption | WRN-40-2 · CIFAR-10-C | classification error: 22.3 -> 18.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- StyleNeophile — Style Neophile: Constantly Seeking Novel Styles for Domain Generalization (CVPR 2022) · better on 4 of 4 shared results
- AugMix — AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty (ICLR 2020) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Generalizing Across Domains via Cross-Gradient Training (ICLR 2018 · similarity 0.91)
- Cross Contrasting Feature Perturbation for Domain Generalization (ICCV 2023 · similarity 0.88)
- Domain Generalization via Entropy Regularization (NeurIPS 2020 · similarity 0.88)
- CrossMatch: Cross-Classifier Consistency Regularization for Open-Set Single Domain Generalization (ICLR 2022 · similarity 0.88)
- DomED: Redesigning Ensemble Distillation for Domain Generalization (ICML 2026 · similarity 0.88)
- Mixup-Induced Domain Extrapolation for Domain Generalization (AAAI 2024 · similarity 0.87)
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.
