eduzhai > Applied Sciences > Engineering >

Cloze Test Helps Effective Video Anomaly Detection via Learning to Complete Video Events

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 9 pages

Abstract: As a vital topic in media content interpretation, video anomaly detection(VAD) has made fruitful progress via deep neural network (DNN). However,existing methods usually follow a reconstruction or frame prediction routine.They suffer from two gaps: (1) They cannot localize video activities in a bothprecise and comprehensive manner. (2) They lack sufficient abilities to utilizehigh-level semantics and temporal context information. Inspired byfrequently-used cloze test in language study, we propose a brand-new VADsolution named Video Event Completion (VEC) to bridge gaps above: First, wepropose a novel pipeline to achieve both precise and comprehensive enclosure ofvideo activities. Appearance and motion are exploited as mutually complimentarycues to localize regions of interest (RoIs). A normalized spatio-temporal cube(STC) is built from each RoI as a video event, which lays the foundation of VECand serves as a basic processing unit. Second, we encourage DNN to capturehigh-level semantics by solving a visual cloze test. To build such a visualcloze test, a certain patch of STC is erased to yield an incomplete event (IE).The DNN learns to restore the original video event from the IE by inferring themissing patch. Third, to incorporate richer motion dynamics, another DNN istrained to infer erased patches optical flow. Finally, two ensemble strategiesusing different types of IE and modalities are proposed to boost VADperformance, so as to fully exploit the temporal context and modalityinformation for VAD. VEC can consistently outperform state-of-the-art methodsby a notable margin (typically 1.5 -5 AUROC) on commonly-used VAD benchmarks.Our codes and results can be verified at this http URL.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×