Presentation 2014-03-13
Violent Scenes Detection in Videos using Automatically Inferred Mid-level Concepts
Shinichi GOTO, Terumasa AOKI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper we propose a novel system for violent scenes detection. Though some previous works have utilized violent concepts annotated by human assessors in order to deal with the ambiguity of the word "violence," our system automatically infers Mid-level Concepts as alternatives. Visual and audio features are extracted at fixed segment-level and used as the input for Multiple Kernel Learning, and finally each violence-score is calculated. We trained the whole system on a dataset from MediaEval 2013 Affect Task and evaluated it by its official metric. The obtained results outperformed the best score in Affect Task.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) violent scenes detection / video processing / machine learning / computer vision / multimedia processing
Paper # PRMU2013-175
Date of Issue

Conference Information
Committee PRMU
Conference Date 2014/3/6(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Violent Scenes Detection in Videos using Automatically Inferred Mid-level Concepts
Sub Title (in English)
Keyword(1) violent scenes detection
Keyword(2) video processing
Keyword(3) machine learning
Keyword(4) computer vision
Keyword(5) multimedia processing
1st Author's Name Shinichi GOTO
1st Author's Affiliation Graduate School of Information Sciences, Tohoku University()
2nd Author's Name Terumasa AOKI
2nd Author's Affiliation New Industry Creation Hatchery Center (NICHe), Tohoku University
Date 2014-03-13
Paper # PRMU2013-175
Volume (vol) vol.113
Number (no) 493
Page pp.pp.-
#Pages 6
Date of Issue