Text-Conditioned Resampler For Long Form Video Understanding

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Korbar, Bruno, Xian, Yongqin, Tonioni, Alessio, Zisserman, Andrew, Tombari, Federico
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916360092647424
author Korbar, Bruno
Xian, Yongqin
Tonioni, Alessio
Zisserman, Andrew
Tombari, Federico
author_facet Korbar, Bruno
Xian, Yongqin
Tonioni, Alessio
Zisserman, Andrew
Tombari, Federico
contents In this paper we present a text-conditioned video resampler (TCR) module that uses a pre-trained and frozen visual encoder and large language model (LLM) to process long video sequences for a task. TCR localises relevant visual features from the video given a text condition and provides them to a LLM to generate a text response. Due to its lightweight design and use of cross-attention, TCR can process more than 100 frames at a time with plain attention and without optimised implementations. We make the following contributions: (i) we design a transformer-based sampling architecture that can process long videos conditioned on a task, together with a training method that enables it to bridge pre-trained visual and language models; (ii) we identify tasks that could benefit from longer video perception; and (iii) we empirically validate its efficacy on a wide variety of evaluation tasks including NextQA, EgoSchema, and the EGO4D-LTA challenge.
format Preprint
id arxiv_https___arxiv_org_abs_2312_11897
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Text-Conditioned Resampler For Long Form Video Understanding
Korbar, Bruno
Xian, Yongqin
Tonioni, Alessio
Zisserman, Andrew
Tombari, Federico
Computer Vision and Pattern Recognition
In this paper we present a text-conditioned video resampler (TCR) module that uses a pre-trained and frozen visual encoder and large language model (LLM) to process long video sequences for a task. TCR localises relevant visual features from the video given a text condition and provides them to a LLM to generate a text response. Due to its lightweight design and use of cross-attention, TCR can process more than 100 frames at a time with plain attention and without optimised implementations. We make the following contributions: (i) we design a transformer-based sampling architecture that can process long videos conditioned on a task, together with a training method that enables it to bridge pre-trained visual and language models; (ii) we identify tasks that could benefit from longer video perception; and (iii) we empirically validate its efficacy on a wide variety of evaluation tasks including NextQA, EgoSchema, and the EGO4D-LTA challenge.
title Text-Conditioned Resampler For Long Form Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2312.11897