Multimodal Violence Detection in Hollywood Movies: State-of-the-Art and Benchmarking

摘要: This chapter introduces a benchmark evaluation targeting the detection of violent scenes in Hollywood movies. The was implemented 2011 and 2012 as an affect task framework international MediaEval initiative. We report on these 2 years evaluation, providing detailed description dataset created, describing state art by studying results achieved participants analysis two best performing multimodal systems. elaborate lessons learned after to provide insights future work emphasizing modeling fusion.