Show simple item record

dc.contributor.authorXu, Mengmeng
dc.contributor.authorPerez-Rua, Juan-Manuel
dc.contributor.authorEscorcia, Victor
dc.contributor.authorMartinez, Brais
dc.contributor.authorZhu, Xiatian
dc.contributor.authorGhanem, Bernard
dc.contributor.authorXiang, Tao
dc.date.accessioned2020-11-25T13:38:16Z
dc.date.available2020-11-25T13:38:16Z
dc.date.issued2020-11-21
dc.identifier.urihttp://hdl.handle.net/10754/666111
dc.description.abstractMany video analysis tasks require temporal localization thus detection of content changes. However, most existing models developed for these tasks are pre-trained on general video action classification tasks. This is because large scale annotation of temporal boundaries in untrimmed videos is expensive. Therefore no suitable datasets exist for temporal boundary-sensitive pre-training. In this paper for the first time, we investigate model pre-training for temporal localization by introducing a novel boundary-sensitive pretext (BSP) task. Instead of relying on costly manual annotations of temporal boundaries, we propose to synthesize temporal boundaries in existing video action classification datasets. With the synthesized boundaries, BSP can be simply conducted via classifying the boundary types. This enables the learning of video representations that are much more transferable to downstream temporal localization tasks. Extensive experiments show that the proposed BSP is superior and complementary to the existing action classification based pre-training counterpart, and achieves new state-of-the-art performance on several temporal localization tasks.
dc.publisherarXiv
dc.relation.urlhttps://arxiv.org/pdf/2011.10830
dc.rightsArchived with thanks to arXiv
dc.titleBoundary-sensitive Pre-training for Temporal Localization in Videos
dc.typePreprint
dc.contributor.departmentElectrical Engineering Program
dc.contributor.departmentElectrical Engineering
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
dc.eprint.versionPre-print
dc.contributor.institutionSamsung AI Centre Cambridge, UK.
dc.contributor.institutionUniversity of Surrey, UK.
dc.identifier.arxivid2011.10830
kaust.personXu, Mengmeng
kaust.personGhanem, Bernard
refterms.dateFOA2020-11-25T13:39:25Z


Files in this item

Thumbnail
Name:
Preprintfile1.pdf
Size:
12.28Mb
Format:
PDF
Description:
Pre-print

This item appears in the following Collection(s)

Show simple item record