AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

MagicFuse: Single Image Fusion for Visual and Semantic Reinforcement

Hao Zhang, Yanping Zha, Zizhuo Li, Meiqi Gong, Jiayi Ma

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multi-modality image fusion · 0 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
fuse diffusion noise per timestep rather than aggregating branch outputs after diffusion (ablation)MagicFuse · MFNetmIoU: 59.74 -> 62.19Improved
fuse diffusion noise per timestep (ablation)MagicFuse · MFNetPSNR: 61.49 -> 63.49Improved
add cross-spectral knowledge generation branch to IKR-only visible enhancement (ablation)MagicFuse · MFNetmIoU: 57.18 -> 62.19Improved
add segmentation head (ablation)MagicFuse · MFNetmIoU: 60.83 -> 62.19Improved
add segmentation head (ablation)MagicFuse · MFNetQabf: 0.50 -> 0.50No clear effect
add one-step estimates to fusion weight estimation (ablation)MagicFuse · MFNetmIoU: 61.05 -> 62.19Improved
add one-step estimates to fusion weight (ablation)MagicFuse · MFNetEN: 7.28 -> 7.29Mixed
tune tau to 0.4 (ablation)MagicFuse · MFNetimage quality and mIoUImproved
apply MagicFuse before SegFormer on CityscapesSegFormer · CityscapesmIoU: 65.80 -> 66.91Improved
compare single-image MagicFuse to multi-modal fusion on FMBMagicFuse · FMBmIoU: 57.72 -> 55.41Hurt
use MagicFuse versus degraded visible imageMagicFuse · MFNetmIoU: 54.09 -> 62.19Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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