Research map / Vision-language and multimodal / Cross-modal learning
Balancing Task-invariant Interaction and Task-specific Adaptation for Unified Image Fusion
Xingyu Hu, Junjun Jiang, Chenyang Wang, Kui Jiang, Xianming Liu, Jiayi Ma
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 |
|---|---|---|---|
| use task-specific objectives instead of unified U2Fusion loss (ablation) | SwinFusion baseline · IVF (LLVIP) | MI: 2.9632 -> 3.6116 | Improved |
| add task-invariant interaction (TI) (ablation) | Baseline-TS · IVF (LLVIP) | MI: 3.6116 -> 3.6853 | Improved |
| add task-specific adaptation (TA) (ablation) | Baseline-TS · IVF (LLVIP) | MI: 3.6116 -> 3.8822 | Improved |
| add FAMO multi-objective optimization (MO) (ablation) | Baseline-TS · IVF (LLVIP) | MI: 3.6116 -> 3.6802 | Improved |
| add MO on top of TI+TA (ablation) | TITA · IVF (LLVIP) | MI: 3.8832 -> 4.1759 | Improved |
| add MO to MEF baseline with TI only (ablation) | Baseline-TS + TI · MEF (MEFB) | MI: 5.2125 -> 5.1837 | Mixed |
| replace SwinFusion with interaction-reduced IrSF (ablation) | SwinFusion · IVF (LLVIP) | MI: 3.6116 -> 3.5334 | Hurt |
| use interaction-enhanced IeSF instead of SF (ablation) | SwinFusion · IVF (LLVIP) | MI: 3.6116 -> 3.6331 | Improved |
| use IPA instead of PA (ablation) | Baseline-TS · IVF (LLVIP) | MI: 3.6694 -> 3.6778 | Improved |
| use IPA instead of Token Exchange (ablation) | Baseline-TS · IVF (LLVIP) | MI: 3.6497 -> 3.6778 | Improved |
| add MUL branch to OAF (ablation) | TITA · IVF (LLVIP) | MI: 3.8994 -> 4.1759 | Improved |
| add dynamic weight prediction to OAF (ablation) | TITA · IVF (LLVIP) | MI: 3.7844 -> 4.1759 | Improved |
Closest papers by meaning
- Task-Customized Mixture of Adapters for General Image Fusion (CVPR 2024 · similarity 0.90)
- One Model for ALL: Low-Level Task Interaction Is a Key to Task-Agnostic Image Fusion (CVPR 2025 · similarity 0.88)
- Task-driven Image Fusion with Learnable Fusion Loss (CVPR 2025 · similarity 0.86)
- Every SAM Drop Counts: Embracing Semantic Priors for Multi-Modality Image Fusion and Beyond (CVPR 2025 · similarity 0.86)
- Beyond Spatial-Domain Supervision: A Relation Constrained Space for Multi-Modal Image Fusion (NeurIPS 2026 · similarity 0.85)
- Towards Minimizing Feature Drift in Model Merging: Layer-wise Task Vector Fusion for Adaptive Knowledge Integration (NeurIPS 2025 · similarity 0.84)
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.
