VISTA: Enhancing Long-Duration and High-Resolution Video Understanding by Video Spatiotemporal Augmentation

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	arXiv.org (Dec 1, 2024), p. n/a
1. Verfasser:	Ren, Weiming
Weitere Verfasser:	Yang, Huan, Min, Jie, Cong, Wei, Chen, Wenhu
Veröffentlicht:	Cornell University Library, arXiv.org
Schlagworte:	Datasets Data augmentation Video data Synthesis Spatiotemporal data Benchmarks High resolution Effectiveness
Online-Zugang:	Citation/Abstract Full text outside of ProQuest
Tags:	Tag hinzufügen Keine Tags, Fügen Sie das erste Tag hinzu!

Beschreibung
Abstract:	Current large multimodal models (LMMs) face significant challenges in processing and comprehending long-duration or high-resolution videos, which is mainly due to the lack of high-quality datasets. To address this issue from a data-centric perspective, we propose VISTA, a simple yet effective Video Spatiotemporal Augmentation framework that synthesizes long-duration and high-resolution video instruction-following pairs from existing video-caption datasets. VISTA spatially and temporally combines videos to create new synthetic videos with extended durations and enhanced resolutions, and subsequently produces question-answer pairs pertaining to these newly synthesized videos. Based on this paradigm, we develop seven video augmentation methods and curate VISTA-400K, a video instruction-following dataset aimed at enhancing long-duration and high-resolution video understanding. Finetuning various video LMMs on our data resulted in an average improvement of 3.3% across four challenging benchmarks for long-video understanding. Furthermore, we introduce the first comprehensive high-resolution video understanding benchmark HRVideoBench, on which our finetuned models achieve a 6.5% performance gain. These results highlight the effectiveness of our framework.
ISSN:	2331-8422
Quelle:	Engineering Database