Papers › StitchFusion: Weaving Any Visual Modalities to Enhance Multimodal Semantic Segmentation

StitchFusion: Weaving Any Visual Modalities to Enhance Multimodal Semantic Segmentation

2 Aug 2024arXiv:2408.01343archive 2025-07-28

Multimodal semantic segmentation shows significant potential for enhancing segmentation accuracy in complex scenes. However, current methods often incorporate specialized feature fusion modules tailored to specific modalities, thereby restricting input flexibility and increasing the number of training parameters. To address these challenges, we propose StitchFusion, a straightforward yet effective modal fusion framework that integrates large-scale pre-trained models directly as encoders and feature fusers. This approach facilitates comprehensive multi-modal and multi-scale feature fusion, accommodating any visual modal inputs. Specifically, Our framework achieves modal integration during encoding by sharing multi-modal visual information. To enhance information exchange across modalities, we introduce a multi-directional adapter module (MultiAdapter) to enable cross-modal information transfer during encoding. By leveraging MultiAdapter to propagate multi-scale information across pre-trained encoders during the encoding process, StitchFusion achieves multi-modal visual information integration during encoding. Extensive comparative experiments demonstrate that our model achieves state-of-the-art performance on four multi-modal segmentation datasets with minimal additional parameters. Furthermore, the experimental integration of MultiAdapter with existing Feature Fusion Modules (FFMs) highlights their complementary nature. Our code is available at StitchFusion_repo.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

SegmentationSemantic SegmentationThermal Image Segmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Semantic Segmentation DeLiVER StitchFusion(RGB-D-E-LiDAR) mIoU 68.18 #2 of 26 Archive leaderboard report
Semantic Segmentation DeLiVER StitchFusion (RGB-D-LiDAR) mIoU 66.65 #4 of 26 Archive leaderboard report
Semantic Segmentation DeLiVER StitchFusion (RGB-D-Event) mIoU 66.03 #6 of 26 Archive leaderboard report
Semantic Segmentation DeLiVER StitchFusion (RGB-Depth) mIoU 65.75 #7 of 26 Archive leaderboard report
Semantic Segmentation DeLiVER StitchFusion (RGB-LiDAR) mIoU 58.03 #13 of 26 Archive leaderboard report
Semantic Segmentation DeLiVER StitchFusion (RGB-Event) mIoU 57.44 #14 of 26 Archive leaderboard report
Semantic Segmentation FMB Dataset StitchFusion+FFMs (RGB-Infrared) mIoU 64.32 #3 of 14 Archive leaderboard report
Semantic Segmentation FMB Dataset StitchFusion (RGB-Infrared) mIoU 63.30 #4 of 14 Archive leaderboard report
Semantic Segmentation MCubeS StitchFusion (RGB-A-D-N) mIoU 53.92 #1 of 22 Archive leaderboard report
Semantic Segmentation MCubeS StitchFusion (RGB-A-D) mIoU 53.26 #2 of 22 Archive leaderboard report
Semantic Segmentation MCubeS StitchFusion (RGB-N) mIoU 53.21 #3 of 22 Archive leaderboard report
Semantic Segmentation MCubeS StitchFusion (RGB-D) mIoU 52.72 #6 of 22 Archive leaderboard report
Semantic Segmentation MCubeS StitchFusion (RGB-A) mIoU 52.68 #7 of 22 Archive leaderboard report
Thermal Image Segmentation MFN Dataset StitchFusion mIOU 58.13 #16 of 55 Archive leaderboard report
Thermal Image Segmentation PST900 StitchFusion (RGB-T) mIoU 85.35 #9 of 22 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Adapter

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections