We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: VLG-Net: Video-Language Graph Matching Network for Video Grounding

Abstract: Grounding language queries in videos aims at identifying the time interval (or moment) semantically relevant to a language query. The solution to this challenging task demands the understanding of videos' and queries' semantic content and the fine-grained reasoning about their multi-modal interactions. Our key idea is to recast this challenge into an algorithmic graph matching problem. Fueled by recent advances in Graph Neural Networks, we propose to leverage Graph Convolutional Networks to model video and textual information as well as their semantic alignment. To enable the mutual exchange of information across the domains, we design a novel Video-Language Graph Matching Network (VLG-Net) to match video and query graphs. Core ingredients include representation graphs, built on top of video snippets and query tokens separately, which are used for modeling the intra-modality relationships. A Graph Matching layer is adopted for cross-modal context modeling and multi-modal fusion. Finally, moment candidates are created using masked moment attention pooling by fusing the moment's enriched snippet features. We demonstrate superior performance over state-of-the-art grounding methods on three widely used datasets for temporal localization of moments in videos with natural language queries: ActivityNet-Captions, TACoS, and DiDeMo.
Comments: 12 pages, 5 figures
Subjects: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
Cite as: arXiv:2011.10132 [cs.CV]
  (or arXiv:2011.10132v1 [cs.CV] for this version)

Submission history

From: Sally Sisi Qu [view email]
[v1] Thu, 19 Nov 2020 22:32:03 GMT (16459kb,D)

Link back to: arXiv, form interface, contact.