VideoMamba: State Space Model for Efficient Video Understanding

TL;DR

VideoMamba introduces a scalable, efficient state space model for video understanding that overcomes limitations of existing methods, enabling superior long-term and multi-modal video analysis without extensive pretraining.

Contribution

It adapts the Mamba model to video, providing a linear-complexity operator for efficient long-term modeling and a novel self-distillation technique for scalability.

Findings

Demonstrates superior long-term video understanding performance.

Achieves high sensitivity in recognizing short-term actions.

Shows robustness in multi-modal video analysis.

Abstract

Addressing the dual challenges of local redundancy and global dependencies in video understanding, this work innovatively adapts the Mamba to the video domain. The proposed VideoMamba overcomes the limitations of existing 3D convolution neural networks and video transformers. Its linear-complexity operator enables efficient long-term modeling, which is crucial for high-resolution long video understanding. Extensive evaluations reveal VideoMamba's four core abilities: (1) Scalability in the visual domain without extensive dataset pretraining, thanks to a novel self-distillation technique; (2) Sensitivity for recognizing short-term actions even with fine-grained motion differences; (3) Superiority in long-term video understanding, showcasing significant advancements over traditional feature-based models; and (4) Compatibility with other modalities, demonstrating robustness in multi-modal contexts. Through these distinct advantages, VideoMamba sets a new benchmark for video understanding, offering a scalable and efficient solution for comprehensive video understanding. All the code and models are available at https://github.com/OpenGVLab/VideoMamba.