Real time violence detection in video

Research output: Contribution to conferencePaperpeer-review

1 Scopus citations


We propose a Violent Flow (ViF) variation using Horn-Schunck instead of Iterative Reweighted Least Squares (IRLS) as optical flow algorithm and Support Vector Machine (SVM) as a classifier, we evaluated the performance in different datasets. In Hockey dataset Horn-Schunck had a better performance, but in Movies and Crowded datasets IRLS was better. We also evaluated the computational cost processing two frames, Horn-Schunck had a very low cost of 0.25 seconds against 7.80 seconds of IRLS enabling its use in real time.

Original languageEnglish
StatePublished - 2016
EventInternational Conference on Pattern Recognition Systems, ICPRS 2016 - Talca, Chile
Duration: 20 Apr 201622 Apr 2016


ConferenceInternational Conference on Pattern Recognition Systems, ICPRS 2016


  • MoSIFT
  • SIFT
  • STIP
  • ViF


Dive into the research topics of 'Real time violence detection in video'. Together they form a unique fingerprint.

Cite this