Real time social distancing violation detection using SVM classifier and deep neural network learning techniques
Authors
Chaudhary, Aatiya
Issue Date
2020
Degree
MSc in Data Analytics
Publisher
Dublin Business School
Rights holder
Rights
Items in eSource are protected by copyright. Previously published items are made available in accordance with the copyright policy of the publisher/copyright holder.
Abstract
Whenever world is hit by any pandemic, it creates immediate global health crises due its deadly spread. Social Distancing is the key measure to control the spread of any pandemic in absence of immediate pharmaceutical inventions. Motivated by current world situation due to COVID-19, the research proposes real time detection framework for monitoring social distancing in public areas for current and any future situations. The detection framework is developed using Support Vector Classifier (SVM) with Histogram of Gradient (HOG) and pre-trained neural network models. The result of two models is compared in terms of dimensions of detection box area. The pre-trained neural network model trained using transfer learning is observed to provide better detection results. It is used to feed on real time video to compute the pairwise centroid distance of the features in three dimensional spaces using tensor flow package. The social distancing violation term is proposed to run the experimental analysis and displays the top view of the detection to give real-time view for security surveillance representing the colour change depicting alerts.