A review of bloody violence in video classification
- 1 October 2017
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 86-91
Abstract
The bloody violence scene in the video will cause different levels of visual and spiritual impact, affecting the health education of the minor children, so the violent scene detection is important for video scientific management and the minor protection. Since European MediaEval proposed Violent Scene Detection (VSD) competition in 2011, the global research scholars actively involved, put forward a number of practical violent video detection methods. This paper reviews the existing methods of violent video detection, analyzes from three main aspects including feature extraction, coding fusion and classifier design. In addition, we also expounds the performance of multimodal features and the relationship, the adaptability of coding fusion method to different features, and also contrast of Softmax classifier and SVM classifier performance respectively. Finally, we give the personal ideas for the research status after summarizing the existing methods.Keywords
This publication has 13 references indexed in Scilit:
- Violent Scene Detection Using Convolutional Neural Networks and Deep Audio Features2016
- Computational optimization for violent scenes detection2016
- Multi-Stream Multi-Class Fusion of Deep Networks for Video Classification2016
- Evaluation of multiple features for violent scenes detectionMultimedia Tools and Applications, 2016
- A modified vector of locally aggregated descriptors approach for fast video classificationMultimedia Tools and Applications, 2015
- Bootstrapping Visual Categorization With Relevant NegativesIEEE Transactions on Multimedia, 2013
- Understanding how Deep Belief Networks perform acoustic modelling2012
- Audio-Visual Fusion for Detecting Violent Scenes in VideosLecture Notes in Computer Science, 2010
- Weakly-Supervised Violence Detection in Movies with Audio and Video Based Co-training2009
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998