Deep Learning Approach for Violence Detection in Urban Areas

Today modern cities tend to grow rapidly. The increased population density brings new challenges in term of public safety. Crime and violence are hard to be detected and managed especially in specific crowd environments like music concerts, sport events or public meetings. To overcome this issue the city administration should implement monitoring systems capable of detecting and analysing such situations. The work presented here combines two approaches that enable implementation of an efficient solution adapted for this purpose. The first one involves sensor networks that prove to be cost effective solution in a smart city environment. They can benefit on the existing surveillance infrastructure and allows rapid deployment. The second approach uses deep learning techniques. They demonstrate outstanding performances in image and actions classification based on a prior learning process. By combining these two approaches we succeed to obtain a real-time and cost-effective solution designed for urban area surveillance networks.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.