Understanding Long Videos with Multimodal Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ranasinghe, Kanchana, Li, Xiang, Kahatapitiya, Kumara, Ryoo, Michael S.
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910999220584448
author Ranasinghe, Kanchana
Li, Xiang
Kahatapitiya, Kumara
Ryoo, Michael S.
author_facet Ranasinghe, Kanchana
Li, Xiang
Kahatapitiya, Kumara
Ryoo, Michael S.
contents Large Language Models (LLMs) have allowed recent LLM-based approaches to achieve excellent performance on long-video understanding benchmarks. We investigate how extensive world knowledge and strong reasoning skills of underlying LLMs influence this strong performance. Surprisingly, we discover that LLM-based approaches can yield surprisingly good accuracy on long-video tasks with limited video information, sometimes even with no video specific information. Building on this, we explore injecting video-specific information into an LLM-based framework. We utilize off-the-shelf vision tools to extract three object-centric information modalities from videos, and then leverage natural language as a medium for fusing this information. Our resulting Multimodal Video Understanding (MVU) framework demonstrates state-of-the-art performance across multiple video understanding benchmarks. Strong performance also on robotics domain tasks establish its strong generality. Code: https://github.com/kahnchana/mvu
format Preprint
id arxiv_https___arxiv_org_abs_2403_16998
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Understanding Long Videos with Multimodal Language Models
Ranasinghe, Kanchana
Li, Xiang
Kahatapitiya, Kumara
Ryoo, Michael S.
Computer Vision and Pattern Recognition
Large Language Models (LLMs) have allowed recent LLM-based approaches to achieve excellent performance on long-video understanding benchmarks. We investigate how extensive world knowledge and strong reasoning skills of underlying LLMs influence this strong performance. Surprisingly, we discover that LLM-based approaches can yield surprisingly good accuracy on long-video tasks with limited video information, sometimes even with no video specific information. Building on this, we explore injecting video-specific information into an LLM-based framework. We utilize off-the-shelf vision tools to extract three object-centric information modalities from videos, and then leverage natural language as a medium for fusing this information. Our resulting Multimodal Video Understanding (MVU) framework demonstrates state-of-the-art performance across multiple video understanding benchmarks. Strong performance also on robotics domain tasks establish its strong generality. Code: https://github.com/kahnchana/mvu
title Understanding Long Videos with Multimodal Language Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.16998