LinVT: Empower Your Image-level Large Language Model to Understand Videos

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gao, Lishuai, Zhong, Yujie, Zeng, Yingsen, Tan, Haoxian, Li, Dengjie, Zhao, Zheng
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910738902155264
author Gao, Lishuai
Zhong, Yujie
Zeng, Yingsen
Tan, Haoxian
Li, Dengjie
Zhao, Zheng
author_facet Gao, Lishuai
Zhong, Yujie
Zeng, Yingsen
Tan, Haoxian
Li, Dengjie
Zhao, Zheng
contents Large Language Models (LLMs) have been widely used in various tasks, motivating us to develop an LLM-based assistant for videos. Instead of training from scratch, we propose a module to transform arbitrary well-trained image-based LLMs into video-LLMs (after being trained on video data). To better adapt image-LLMs for processing videos, we introduce two design principles: linear transformation to preserve the original visual-language alignment and representative information condensation from redundant video content. Guided by these principles, we propose a plug-and-play Linear Video Tokenizer(LinVT), which enables existing image-LLMs to understand videos. We benchmark LinVT with six recent visual LLMs: Aquila, Blip-3, InternVL2, Mipha, Molmo and Qwen2-VL, showcasing the high compatibility of LinVT. LinVT-based LLMs achieve state-of-the-art performance across various video benchmarks, illustrating the effectiveness of LinVT in multi-modal video understanding.
format Preprint
id arxiv_https___arxiv_org_abs_2412_05185
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LinVT: Empower Your Image-level Large Language Model to Understand Videos
Gao, Lishuai
Zhong, Yujie
Zeng, Yingsen
Tan, Haoxian
Li, Dengjie
Zhao, Zheng
Computer Vision and Pattern Recognition
Machine Learning
Multimedia
Large Language Models (LLMs) have been widely used in various tasks, motivating us to develop an LLM-based assistant for videos. Instead of training from scratch, we propose a module to transform arbitrary well-trained image-based LLMs into video-LLMs (after being trained on video data). To better adapt image-LLMs for processing videos, we introduce two design principles: linear transformation to preserve the original visual-language alignment and representative information condensation from redundant video content. Guided by these principles, we propose a plug-and-play Linear Video Tokenizer(LinVT), which enables existing image-LLMs to understand videos. We benchmark LinVT with six recent visual LLMs: Aquila, Blip-3, InternVL2, Mipha, Molmo and Qwen2-VL, showcasing the high compatibility of LinVT. LinVT-based LLMs achieve state-of-the-art performance across various video benchmarks, illustrating the effectiveness of LinVT in multi-modal video understanding.
title LinVT: Empower Your Image-level Large Language Model to Understand Videos
topic Computer Vision and Pattern Recognition
Machine Learning
Multimedia
url https://arxiv.org/abs/2412.05185