IQViC: In-context, Question Adaptive Vision Compressor for Long-term Video Understanding LMMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yamao, Sosuke, Miyahara, Natsuki, Harazono, Yuki, Takeuchi, Shun
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929632283983872
author Yamao, Sosuke
Miyahara, Natsuki
Harazono, Yuki
Takeuchi, Shun
author_facet Yamao, Sosuke
Miyahara, Natsuki
Harazono, Yuki
Takeuchi, Shun
contents With the increasing complexity of video data and the need for more efficient long-term temporal understanding, existing long-term video understanding methods often fail to accurately capture and analyze extended video sequences. These methods typically struggle to maintain performance over longer durations and to handle the intricate dependencies within the video content. To address these limitations, we propose a simple yet effective large multi-modal model framework for long-term video understanding that incorporates a novel visual compressor, the In-context, Question Adaptive Visual Compressor (IQViC). The key idea, inspired by humans' selective attention and in-context memory mechanisms, is to introduce a novel visual compressor and incorporate efficient memory management techniques to enhance long-term video question answering. Our framework utilizes IQViC, a transformer-based visual compressor, enabling question-conditioned in-context compression, unlike existing methods that rely on full video visual features. This selectively extracts relevant information, significantly reducing memory token requirements. Through extensive experiments on a new dataset based on InfiniBench for long-term video understanding, and standard benchmarks used for existing methods' evaluation, we demonstrate the effectiveness of our proposed IQViC framework and its superiority over state-of-the-art methods in terms of video understanding accuracy and memory efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2412_09907
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle IQViC: In-context, Question Adaptive Vision Compressor for Long-term Video Understanding LMMs
Yamao, Sosuke
Miyahara, Natsuki
Harazono, Yuki
Takeuchi, Shun
Computer Vision and Pattern Recognition
With the increasing complexity of video data and the need for more efficient long-term temporal understanding, existing long-term video understanding methods often fail to accurately capture and analyze extended video sequences. These methods typically struggle to maintain performance over longer durations and to handle the intricate dependencies within the video content. To address these limitations, we propose a simple yet effective large multi-modal model framework for long-term video understanding that incorporates a novel visual compressor, the In-context, Question Adaptive Visual Compressor (IQViC). The key idea, inspired by humans' selective attention and in-context memory mechanisms, is to introduce a novel visual compressor and incorporate efficient memory management techniques to enhance long-term video question answering. Our framework utilizes IQViC, a transformer-based visual compressor, enabling question-conditioned in-context compression, unlike existing methods that rely on full video visual features. This selectively extracts relevant information, significantly reducing memory token requirements. Through extensive experiments on a new dataset based on InfiniBench for long-term video understanding, and standard benchmarks used for existing methods' evaluation, we demonstrate the effectiveness of our proposed IQViC framework and its superiority over state-of-the-art methods in terms of video understanding accuracy and memory efficiency.
title IQViC: In-context, Question Adaptive Vision Compressor for Long-term Video Understanding LMMs
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.09907