Complex event detection using semantic saliency and nearly-isotonic SVM

Xiaojun Chang, Yi Yang, Eric P. Xing, Yao-Liang Yu

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

57 Citations (Scopus)


We aim to detect complex events in long Internet videos that may last for hours. A major challenge in this setting is that only a few shots in a long video are relevant to the event of interest while others are irrelevant or even misleading. Instead of indifferently pooling the shots, we first define a novel notion of semantic saliency that assesses the relevance of each shot with the event of interest. We then prioritize the shots according to their saliency scores since shots that are semantically more salient are expected to contribute more to the final event detector. Next, we propose a new isotonic regularizer that is able to exploit the semantic ordering information. The resulting nearly-isotonic SVM classifier exhibits higher discriminative power. Computationally, we develop an efficient implementation using the proximal gradient algorithm, and we prove new, closed-form proximal steps. We conduct extensive experiments on three real-world video datasets and confirm the effectiveness of the proposed approach.
Original languageEnglish
Title of host publicationProceedings of the 32nd International Conference on Machine Learning
Subtitle of host publicationLille, France — July 06 - 11, 2015
EditorsFrancis Bach, David Blei
Place of PublicationRed Hook NY USA
PublisherProceedings of Machine Learning Research (PMLR)
Number of pages10
Publication statusPublished - 2015
Externally publishedYes
EventInternational Conference on Machine Learning 2015 - Lille Grand Palais, Lille, France
Duration: 6 Jul 201511 Jul 2015
Conference number: 32nd

Publication series

NameProceedings of Machine Learning Research
PublisherInternational Conference on Machine Learning
ISSN (Electronic)2640-3498


ConferenceInternational Conference on Machine Learning 2015
Abbreviated titleICML 2015
Internet address

Cite this