AI Paper Atlas

Research map / Detection and segmentation / CNN features

Multi-interactive Feature Learning and a Full-time Multi-modality Benchmark for Image Fusion and Segmentation

Jinyuan Liu, Zhu Liu, Guanyao Wu, Long Ma, Risheng Liu, Wei Zhong, Zhongxuan Luo, Xin Fan

ICCV 2023 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 375 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
replace direct feature aggregation (concatenate) with full hierarchical interactive attention (ablation)SegMiF · MFNet segmentationmIoU: 52.7 -> 56.1Improved
replace direct feature aggregation (concatenate) with full HIA (ablation)SegMiF · FMB segmentationmIoU: 51.3 -> 54.8Improved
add semantic-oriented attention module (SoAM) to HIA (ablation)SegMiF HIA · FMB segmentationmIoU: 52.3 -> 54.8Improved
add modality-oriented attention module (MoAM) to HIA (ablation)SegMiF HIA · MFNet segmentationmIoU: 53.4 -> 56.1Improved
use attention variants M1 (neither SoAM nor MoAM) versus aggregation (ablation)SegMiF HIA · MFNet segmentationmIoU: 52.7 -> 51.9No clear effect
use dynamic weighting factor instead of other multi-task strategies (ablation)SegMiF training · MFNet and FMBfusion and segmentation qualityImproved
joint fusion and segmentation with interactive features versus best prior methodSegMiF · FMB segmentationmIoU: 50.9 -> 54.8Improved
SegMiF versus dual-stream and fusion baselines on mAccSegMiF · MFNet segmentationmAcc: 75.4 (LASNet) vs 74.8Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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