Research map / Label-efficient and robust learning / Domain adaptation and generalization
Deeper, Broader and Artier Domain Generalization
Da Li, Yongxin Yang, Yi-Zhe Song, Timothy M. Hospedales
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 2-hot tensor weight generation on final layer (ablation) | Ours-Full (low-rank domain-generalization CNN) · PACS | average accuracy (%): 62.30 -> 62.62 | Improved |
| add low-rank decomposition of final layer (ablation) | Ours-Full (low-rank domain-generalization CNN) · PACS | average accuracy (%): 62.62 -> 64.10 | Improved |
| apply the DG strategy at every layer instead of the last only (ablation) | Ours-Full (low-rank domain-generalization CNN) · PACS | average accuracy (%): 64.10 -> 69.21 | Improved |
| use low-rank DG model instead of 1HNN with same features | Ours-Full (low-rank domain-generalization CNN) · PACS | average accuracy (%): 64.29 -> 64.65 | Improved |
| use low-rank DG model instead of 1HNN with same features | Ours-Full (low-rank domain-generalization CNN) · VLCS | average accuracy (%): 65.46 -> 69.65 | Improved |
| replace Deep-All fine-tuning with Ours-Full | Ours-Full (low-rank domain-generalization CNN) · PACS | average accuracy (%): 67.05 -> 69.21 | Improved |
| replace Deep-All fine-tuning with Ours-Full | Ours-Full (low-rank domain-generalization CNN) · VLCS | average accuracy (%): 72.02 -> 72.11 | No clear effect |
| replace Deep-All fine-tuning with Ours-Full | Ours-Full (low-rank domain-generalization CNN) · VLCS (unseen Sun) | accuracy (%): 64.16 -> 61.32 | Hurt |
| replace Deep-All fine-tuning with Ours-Full | Ours-Full (low-rank domain-generalization CNN) · PACS (unseen Art painting) | accuracy (%): 63.30 -> 62.86 | Hurt |
| replace Undo-Bias with Ours-MLP on SIFT-BOW features | Ours-Full (low-rank domain-generalization CNN) · VLCS mean AP, Car | average precision (%): 60.28 -> 62.37 | Improved |
| replace Undo-Bias with Ours-MLP on SIFT-BOW features | Ours-Full (low-rank domain-generalization CNN) · VLCS mean AP, Bird | average precision (%): 15.29 -> 13.72 | Hurt |
| fine-tune a CNN on aggregated sources (Deep-All) instead of prior DG methods on DeCAF6 | Ours-Full (low-rank domain-generalization CNN) · VLCS | average accuracy (%): 68.60 -> 72.02 | Improved |
Closest papers by meaning
- PracticalDG: Perturbation Distillation on Vision-Language Models for Hybrid Domain Generalization (CVPR 2024 · similarity 0.90)
- Is Large-scale Pretraining the Secret to Good Domain Generalization? (ICLR 2025 · similarity 0.90)
- Domain Generalization in CLIP via Learning with Diverse Text Prompts (CVPR 2025 · similarity 0.90)
- Multimodal Unsupervised Domain Generalization by Retrieving Across the Modality Gap (ICLR 2025 · similarity 0.90)
- Robust Domain Generalization under Divergent Marginal and Conditional Distributions (NeurIPS 2026 · similarity 0.89)
- START: A Generalized State Space Model with Saliency-Driven Token-Aware Transformation (NeurIPS 2024 · 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.
