Two-person interaction recognition via spatial multiple instance embedding

Simple item page
dc.citation.epage73en_US
dc.citation.spage63en_US
dc.citation.volumeNumber32en_US
dc.contributor.authorSener F.en_US
dc.contributor.authorIkizler-Cinbis, N.en_US
dc.date.accessioned2016-02-08T09:33:59Z
dc.date.available2016-02-08T09:33:59Z
dc.date.issued2015en_US
dc.departmentDepartment of Computer Engineeringen_US
dc.description.abstractAbstract In this work, we look into the problem of recognizing two-person interactions in videos. Our method integrates multiple visual features in a weakly supervised manner by utilizing an embedding-based multiple instance learning framework. In our proposed method, first, several visual features that capture the shape and motion of the interacting people are extracted from each detected person region in a video. Then, two-person visual descriptors are formed. Since the relative spatial locations of interacting people are likely to complement the visual descriptors, we propose to use spatial multiple instance embedding, which implicitly incorporates the distances between people into the multiple instance learning process. Experimental results on two benchmark datasets validate that using two-person visual descriptors together with spatial multiple instance learning offers an effective way for inferring the type of the interaction. © 2015 Elsevier Inc.en_US
dc.identifier.doi10.1016/j.jvcir.2015.07.016en_US
dc.identifier.issn10473203
dc.identifier.urihttp://hdl.handle.net/11693/20726
dc.language.isoEnglishen_US
dc.publisherAcademic Press Inc.en_US
dc.relation.isversionofhttp://dx.doi.org/10.1016/j.jvcir.2015.07.016en_US
dc.source.titleJournal of Visual Communication and Image Representationen_US
dc.subjectActivity recognitionen_US
dc.subjectHuman actionsen_US
dc.subjectHuman interaction recognitionen_US
dc.subjectHuman interactionsen_US
dc.subjectMultiple instance learningen_US
dc.subjectSpatial embeddingen_US
dc.subjectVideo analysisen_US
dc.subjectVideo retrievalen_US
dc.subjectImage recognitionen_US
dc.subjectMotion estimationen_US
dc.subjectActivity recognitionen_US
dc.subjectHuman actionsen_US
dc.subjectHuman interaction recognitionen_US
dc.subjectHuman interactionsen_US
dc.subjectMultiple instance learningen_US
dc.subjectSpatial embeddingen_US
dc.subjectVideo analysisen_US
dc.subjectVideo retrievalen_US
dc.subjectLearning systemsen_US
dc.titleTwo-person interaction recognition via spatial multiple instance embeddingen_US
dc.typeArticleen_US

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Two-person interaction recognition via spatial multiple instance embedding.pdf
Size:
3.18 MB
Format:
Adobe Portable Document Format
Description:
Full printable version
Download

Collections

Department of Computer Engineering