AI RESEARCH
DGSSM: Diffusion guided state-space models for multimodal salient object detection
arXiv CS.LG
•
ArXi:2604.17585v1 Announce Type: cross Salient object detection (SOD) requires modeling both long-range contextual dependencies and fine-grained structural details, which remains challenging for convolutional, transformer-based, and Mamba-based state space models. While recent Mamba-based state space approaches enable efficient global reasoning, they often struggle to recover precise object boundaries.