MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding

TL;DR

This paper introduces MA-LMM, a memory-augmented multimodal model that enables long-term video understanding by storing and referencing historical video data, overcoming the limitations of existing models in processing extended video sequences.

Contribution

The paper presents a novel memory bank mechanism integrated into multimodal LLMs, allowing efficient long-term video analysis without exceeding context or memory limits.

Findings

Achieves state-of-the-art results on multiple long-video understanding datasets

Effectively handles long-term video question answering and captioning tasks

Demonstrates seamless integration of memory bank into existing multimodal models

Abstract

With the success of large language models (LLMs), integrating the vision model into LLMs to build vision-language foundation models has gained much more interest recently. However, existing LLM-based large multimodal models (e.g., Video-LLaMA, VideoChat) can only take in a limited number of frames for short video understanding. In this study, we mainly focus on designing an efficient and effective model for long-term video understanding. Instead of trying to process more frames simultaneously like most existing work, we propose to process videos in an online manner and store past video information in a memory bank. This allows our model to reference historical video content for long-term analysis without exceeding LLMs' context length constraints or GPU memory limits. Our memory bank can be seamlessly integrated into current multimodal LLMs in an off-the-shelf manner. We conduct extensive experiments on various video understanding tasks, such as long-video understanding, video question answering, and video captioning, and our model can achieve state-of-the-art performances across multiple datasets. Code available at https://boheumd.github.io/MA-LMM/.