VMamba: Visual State Space Model

TL;DR

VMamba introduces a novel vision backbone using Visual State-Space blocks with linear time complexity, effectively capturing contextual information in 2D vision data, and demonstrates superior efficiency and performance across various visual tasks.

Contribution

It adapts the Mamba state-space model into a vision-specific architecture with a new SS2D module, achieving efficient processing of 2D visual data.

Findings

Superior input scaling efficiency compared to benchmarks

Effective collection of contextual information from multiple perspectives

Promising performance across diverse visual perception tasks

Abstract

Designing computationally efficient network architectures remains an ongoing necessity in computer vision. In this paper, we adapt Mamba, a state-space language model, into VMamba, a vision backbone with linear time complexity. At the core of VMamba is a stack of Visual State-Space (VSS) blocks with the 2D Selective Scan (SS2D) module. By traversing along four scanning routes, SS2D bridges the gap between the ordered nature of 1D selective scan and the non-sequential structure of 2D vision data, which facilitates the collection of contextual information from various sources and perspectives. Based on the VSS blocks, we develop a family of VMamba architectures and accelerate them through a succession of architectural and implementation enhancements. Extensive experiments demonstrate VMamba's promising performance across diverse visual perception tasks, highlighting its superior input scaling efficiency compared to existing benchmark models. Source code is available at https://github.com/MzeroMiko/VMamba.