AI Paper Atlas

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

AAAI 2026 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Cross-domain few-shot with transformers · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
apply SWIFT to a source-pretrained model for cross-sensor adaptationFusionNet, QB->GF2 · QB->GF2 full-resolutionHQNR: 0.840 -> 0.934Improved
apply SWIFT to ADWM for cross-sensor adaptationADWM, QB->GF2 · QB->GF2 full-resolutionHQNR: 0.728 -> 0.913Improved
apply SWIFT instead of retraining on target domainSSDiff · QB->GF2 full-resolutionadaptation time (s): 32336.5 -> 1077.9Improved
apply SWIFT to WFANetWFANet, QB->GF2 · QB->GF2 full-resolutionHQNR: 0.900 -> 0.920Improved
apply SWIFT to ADWMADWM, GF2->QB · GF2->QB full-resolutionHQNR: 0.516 -> 0.769Improved
apply SWIFT to a model transferred to a new sensorFusionNet, GF2->QB · GF2->QB full-resolutionHQNR: 0.557 -> 0.794 (retrained model: 0.890)Mixed
apply SWIFT on WV3->WV2 transferU2Net, WV3->WV2 · WV3->WV2 full-resolutionHQNR: 0.87 -> 0.94Improved
apply SWIFT on WV3->WV2 transfer, reduced resolutionFusionNet, WV3->WV2 · WV3->WV2 reduced-resolutionERGAS: 5.04 -> 2.68Improved
use density-aware farthest point sampling instead of random sampling (ablation)SWIFT-FusionNet, GF2->QB · GF2->QB full-resolution, 3% samplesHQNR: 0.75 -> 0.79Improved
use density-aware farthest point sampling instead of random sampling (ablation)SWIFT-FusionNet, GF2->QB · GF2->QB full-resolution, 3% samplesMMD: 0.053 -> 0.044Improved
use density-aware farthest point sampling instead of random sampling (ablation)SWIFT-FusionNet, GF2->QB · GF2->QB full-resolution, 3% samplesDλ: 0.14 -> 0.17Hurt
use dynamic parameter selection instead of fixed-ratio selection (ablation)SWIFT-FusionNet, QB · GF2->QB full-resolutionHQNR: 34.1% of parameters surpasses any fixed ratioImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy