Automatic Detection of Riots Using Deep Learning

EasyChair Preprint 1976, version 1

Abstract

A Riot like situation styles serious consequence on societal and individual security and a simple primary cautioning of any fierce, vehement, vicious, force-full or violent movement could significantly condense these dangers. At present, there be located oodles of video surveillance kit applied in civic places, such as bus-stations, highways, airports, Signal-squares, crossings, and railway_stations. It is of vivacious importance to examine the detrimental anomalous fillings from enormous quantities of surveillance-video data. For the thought of real-world application, this effort focuses on the challenging task of detecting violent situations in videotapes & aims to proposition a fangled way that could automatically distinguish violent behaviours by resources of computer vision methods. We take a proportion of videotapes & we train those precise sequences as violent situation & once we have got a model ready, we deploy for example on intelligent surveillance camera, any action which is close to this precise entity would be classified as violent situation and we can direct an alarm/warning back to the control-room for further necessary steps.

Keyphrases: Artificial Intelligence, Batch Normalization, CNN RNN LSTM, Collective activity recognition, Convolutional Neural Network, Gaussian optical_flow Model, Long Short-Term-storage, Motion SIFT, Non-violence, Optical_flow Orientation Histogram, Oriented Violent Flows, Random Forest, Recurrent_Neural_Network, Riots Detection, Space-Time Interest Points, Spacetime volume, Support_Vector_Machine, activity recognition, bag of visual word, computer vision, deep learning, group activity recognition, high-level, improvised Dense Trajectories, k-nearest neighbors, machine learning, neural network, neural networks, surveillance camera, surveillance video

@booklet{EasyChair:1976,
  author    = {Mayur Jadhav and V. A. Chakkarwar},
  title     = {Automatic Detection of Riots Using Deep Learning},
  howpublished = {EasyChair Preprint 1976},
  year      = {EasyChair, 2019}}