Research map / Label-efficient and robust learning / Domain adaptation and generalization
SWIFT:A General Sensitive Weight Identification Framework for Fast Sensor-Transfer Pansharpening
Zeyu Xia, Chenxi Sun, Tianyu Xin, Yubo Zeng, Haoyu Chen, Liang-Jian Deng
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 |
|---|---|---|---|
| apply SWIFT to a source-pretrained model for cross-sensor adaptation | FusionNet, QB->GF2 · QB->GF2 full-resolution | HQNR: 0.840 -> 0.934 | Improved |
| apply SWIFT to ADWM for cross-sensor adaptation | ADWM, QB->GF2 · QB->GF2 full-resolution | HQNR: 0.728 -> 0.913 | Improved |
| apply SWIFT instead of retraining on target domain | SSDiff · QB->GF2 full-resolution | adaptation time (s): 32336.5 -> 1077.9 | Improved |
| apply SWIFT to WFANet | WFANet, QB->GF2 · QB->GF2 full-resolution | HQNR: 0.900 -> 0.920 | Improved |
| apply SWIFT to ADWM | ADWM, GF2->QB · GF2->QB full-resolution | HQNR: 0.516 -> 0.769 | Improved |
| apply SWIFT to a model transferred to a new sensor | FusionNet, GF2->QB · GF2->QB full-resolution | HQNR: 0.557 -> 0.794 (retrained model: 0.890) | Mixed |
| apply SWIFT on WV3->WV2 transfer | U2Net, WV3->WV2 · WV3->WV2 full-resolution | HQNR: 0.87 -> 0.94 | Improved |
| apply SWIFT on WV3->WV2 transfer, reduced resolution | FusionNet, WV3->WV2 · WV3->WV2 reduced-resolution | ERGAS: 5.04 -> 2.68 | Improved |
| use density-aware farthest point sampling instead of random sampling (ablation) | SWIFT-FusionNet, GF2->QB · GF2->QB full-resolution, 3% samples | HQNR: 0.75 -> 0.79 | Improved |
| use density-aware farthest point sampling instead of random sampling (ablation) | SWIFT-FusionNet, GF2->QB · GF2->QB full-resolution, 3% samples | MMD: 0.053 -> 0.044 | Improved |
| use density-aware farthest point sampling instead of random sampling (ablation) | SWIFT-FusionNet, GF2->QB · GF2->QB full-resolution, 3% samples | Dλ: 0.14 -> 0.17 | Hurt |
| use dynamic parameter selection instead of fixed-ratio selection (ablation) | SWIFT-FusionNet, QB · GF2->QB full-resolution | HQNR: 34.1% of parameters surpasses any fixed ratio | Improved |
Closest papers by meaning
- A General Adaptive Dual-level Weighting Mechanism for Remote Sensing Pansharpening (CVPR 2025 · similarity 0.86)
- PanAdapter: Two-Stage Fine-Tuning with Spatial-Spectral Priors Injecting for Pansharpening (AAAI 2025 · similarity 0.86)
- RECAST: Reparameterized, Compact weight Adaptation for Sequential Tasks (ICLR 2025 · similarity 0.83)
- Implicit Modeling for Transferability Estimation of Vision Foundation Models (NeurIPS 2025 · similarity 0.83)
- A Closer Look at the Few-Shot Adaptation of Large Vision-Language Models (CVPR 2024 · similarity 0.83)
- ModHiFi: Identifying High Fidelity predictive components for Model Modification (NeurIPS 2025 · similarity 0.82)
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.
