AI Paper Atlas

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

Multi-modal Frequency Decomposition Network for Semantic Scene Completion

Die Zuo, Lubo Wang, Ruonan Liu, Qing Guo, Chong Wang, Dongdong Wu, Wei Feng, Kairui Yang, Di Lin

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multi-modality image fusion

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
add MAFF to addition-based fusion (ablation)MFDNet · NYUv2 testmIoU (%): 54.5 -> 55.7Improved
add FDC (ablation)MFDNet · NYUv2 testmIoU (%): 54.5 -> 55.1Improved
add both MAFF and FDC (ablation)MFDNet · NYUv2 testmIoU (%): 54.5 -> 57.0Improved
learn fusion weights in frequency domain instead of spatial domain (ablation)MAFF · NYUv2 testmIoU (%): 55.5 -> 56.1Improved
learn weights from overall feature for each band (overall-to-band) instead of band-to-band (ablation)MAFF · NYUv2 testmIoU (%): 56.5 -> 57.0Improved
spatial overall-to-overall weighting versus no weights (ablation)MAFF · NYUv2 testIoU (%): 76.3 -> 76.0Hurt
add only high-frequency components of shallow features instead of overall features (ablation)FDC · NYUv2 testmIoU (%): 56.1 -> 56.6Improved
add learned weights to high-frequency compensation (ablation)FDC · NYUv2 testmIoU (%): 56.6 -> 57.0Improved
learn FDC weights from coarse feature rather than shallow features (ablation)FDC · NYUv2 testmIoU (%): 56.3 -> 57.0Improved
replace AMMNet modulation fusion with MAFF (both without extras)AMMNet fusion · NYUv2 testmIoU (%): 55.2 -> 55.7Improved
use MAFF fusion instead of AMMNet modulationAMMNet fusion · NYUv2 testparameters (M): 20.85 -> 4.79Improved

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