Multimodal collaborative saliency object detection network using MCSDNet

doi:10.1016/j.displa.2026.103467

Multimodal collaborative saliency object detection network using MCSDNet

2026-04-08

0

PRE

AI

DOI：10.1016/j.displa.2026.103467

Original

Original request for help

Save

Abstract

En 中文

• Novel three-stream architecture optimizing CNN, CapsNet, and boundary-guided network for feature synergy. • Dynamic cross-modal interaction (SIM/CAM) enabling semantic alignment and noise suppression. • Achieved top performance in 13 out of 20 saliency metrics, and ranked first on four datasets for boundary-related metrics, with improvements of 3%, 2.7%, 0.7%, and 2.9% over the second-best method. • Runs at 22.4 FPS on an NVIDIA GeForce RTX 4080 SUPER (16 GB) GPU, ranking third among all compared methods.

Keywords:

Multimodal collaboration

Saliency detection

CNN

CapsNet

Boundary-guided network

Multimodal collaborative saliency object detection network using MCSDNet

Abstract

Journal

Organization