Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Generalization by Solving Jigsaw Puzzles
Check 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 image-level jigsaw prediction as an auxiliary task to object classification | AlexNet JiGen · PACS | average accuracy: 71.52 -> 73.38 | Helped |
| Add image-level jigsaw prediction as an auxiliary task to object classification | AlexNet JiGen · VLCS | average accuracy: 72.66 -> 73.19 | Helped |
| Add image-level jigsaw prediction as an auxiliary task to object classification | ResNet-18 JiGen · Office-Home | average accuracy: 60.51 -> 61.20 | Helped |
| Add image-level jigsaw prediction as an auxiliary task to object classification | ResNet-18 JiGen · PACS | average accuracy: 79.05 -> 80.51 | Mixed |
| Include unlabeled target images in jigsaw training and set target jigsaw weight to 0.3 | JiGen · PACS domain adaptation | average accuracy: 85.30 -> 86.46 | Helped |
| Add the jigsaw task with alpha 0.1 or 1 while keeping beta at 0.6 (ablation) | AlexNet JiGen · PACS | average accuracy | Helped |
| Set beta to 1 while retaining a high jigsaw loss weight, so training uses only ordered images (ablation) | AlexNet JiGen · PACS | average accuracy | Hurt |
| Train with shuffled images only (beta 0) (ablation) | AlexNet JiGen · PACS | accuracy: <20% | Hurt |
| Vary the number of jigsaw classes from 5 to 1000 (ablation) | AlexNet JiGen · PACS | accuracy: 1.5 percentage points | No clear effect |
| Vary the patch grid size from 2×2 to 4×4 (ablation) | AlexNet JiGen · PACS | accuracy | No clear effect |
| Use rotation prediction instead of jigsaw prediction as the auxiliary task (ablation) | AlexNet domain generalization model · PACS | average accuracy: 71.52 -> 72.13 | Helped |
Closest papers by meaning
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 2024 · similarity 0.86)
- Multimodal Unsupervised Domain Generalization by Retrieving Across the Modality Gap (ICLR 2025 · similarity 0.86)
- When Domain Generalization meets Generalized Category Discovery: An Adaptive Task-Arithmetic Driven Approach (CVPR 2025 · similarity 0.86)
- SoMA: Singular Value Decomposed Minor Components Adaptation for Domain Generalizable Representation Learning (CVPR 2025 · similarity 0.85)
- OCRT: Boosting Foundation Models in the Open World with Object-Concept-Relation Triad (CVPR 2025 · similarity 0.85)
- Reasoning-Driven Multimodal LLM for Domain Generalization (ICLR 2026 · similarity 0.85)
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.