Violence Detection of CCTV Footage Using Deep Learning

Date
2020
Research Supervisor
Editor
Journal Title
Volume
Issue
Journal Title
Journal ISSN
Volume Title
Abstract
Use of deep learning in computer vision became popular in recent years. Using this to detect fighting or clashing in public gathering areas and other fight-prone areas, such as prisons, bars, etc. could be useful for the law-enforcement agencies to act quickly and take the situation under control. With this motivation in our mind we developed a model which can detect fighting or clashing in real time by analyzing CCTV footage. Here, we used CNN and LSTM from very basic to advanced architecture. We used 2D and 3D convolution model and analyzed which works better for our dataset. We also used Inception Modules in Time Distributed CNN and LSTM Architecture and then we ensembled our models together in several combinations to see what works better. With our models we were able to predict the data with 83% accuracy. Our dataset consists of two classes. Fighting and non-fighting. With the dataset, we performed binary classification using deep learning
Description
Keywords
Citation
Department Name
Electrical and Computer Engineering
Publisher
North South University
Printed Thesis
DOI
ISSN
ISBN