EntroFormer: An entropy-based sparse vision transformer for real-time semantic segmentation

doi:10.1016/j.cviu.2025.104482

EntroFormer: An entropy-based sparse vision transformer for real-time semantic segmentation

2025-08-20

0

PRE

AI

DOI：10.1016/j.cviu.2025.104482

Original

Original request for help

Save

Abstract

En 中文

• A sparse attention mechanism based on information entropy to enhance image semantic segmentation by focusing on the most informative regions. • An entropy-based attention mechanism is employed to identify the most informative regions for effectively capturing long-range dependencies. • A real-time semantic segmentation network that outperforms existing state-of-the-art methods while maintaining comparable parameter counts and computational cost.

Keywords:

sparse attention

information entropy

image semantic segmentation

long-range dependencies

real-time segmentation

EntroFormer: An entropy-based sparse vision transformer for real-time semantic segmentation

Abstract

Journal

Organization