Snakes and Ladders: Two Steps Up for VideoMamba

TL;DR

This paper introduces VideoMambaPro, an improved video understanding model that addresses limitations of Mamba in vision tasks, achieving higher accuracy and efficiency in video classification benchmarks.

Contribution

It provides a theoretical analysis of Mamba's limitations and proposes VideoMambaPro with novel techniques that outperform previous models on key datasets.

Findings

VideoMambaPro surpasses VideoMamba by 1.6-2.8% on Kinetics-400

VideoMambaPro exceeds VideoMamba by 1.1-1.9% on Something-Something V2

Models perform well even without extensive pre-training

Abstract

Video understanding requires the extraction of rich spatio-temporal representations, which transformer models achieve through self-attention. Unfortunately, self-attention poses a computational burden. In NLP, Mamba has surfaced as an efficient alternative for transformers. However, Mamba's successes do not trivially extend to vision tasks, including those in video analysis. In this paper, we theoretically analyze the differences between self-attention and Mamba. We identify two limitations in Mamba's token processing: historical decay and element contradiction. We propose VideoMambaPro (VMP) that solves the identified limitations by adding masked backward computation and elemental residual connections to a VideoMamba backbone. Differently sized VideoMambaPro models surpass VideoMamba by 1.6-2.8% and 1.1-1.9% top-1 on Kinetics-400 and Something-Something V2, respectively. Even without extensive pre-training, our models present an increasingly attractive and efficient alternative to current transformer models. Moreover, our two solutions are orthogonal to recent advances in Vision Mamba models, and are likely to provide further improvements in future models.