SnAG: Scalable and Accurate Video Grounding

arXiv:2404.02257v1 Announce Type: new
Abstract: Temporal grounding of text descriptions in videos is a central problem in vision-language learning and video understanding. Existing methods often prioritize accuracy over scalability -- they have been optimized for grounding only a few text queries within short videos, and fail to scale up to long videos with hundreds of queries. In this paper, we study the effect of cross-modal fusion on the scalability of video grounding models. Our analysis establishes late fusion as a …

abstract accuracy arxiv cs.cv language queries scalability scalable scale temporal text type understanding video videos video understanding vision