Intelligent Video Analytics for Better Planning of Smart Cities using Edge Computing Savita Shetty 1 , Raghav Maheshwari 2 , Rithika Mehta 3 and Shirsh Vardhan Kashyap 4 1-4 Ramaiah Institute Of Technology , Visvesvaraya Technological University, Belagavi Bangalore, India Email: savita_ks1@msrit.edu, raghav.ddps2@gmail.com, mehtarithika@gmail.com, shirshvardhan@gmail.com Abstract—Recent years have shown how machine learning and deep learning techniques can be used to radically transform smart cities. However, current cloud-centric processing approaches present latency, privacy, and bandwidth-related challenges. Edge computing is a solution to the above challenges where the video feeds captured can be analyzed right at the source of generation of the feed while deeper insights and analytics can still be performed on the cloud. The edge computing approach to video analytics also enables us to take action to certain situations in real-time which is not possible in cloud-centric approaches. In this paper, we propose certain methods for better planning of smart cities using edge computing. The data points generated for the proposed methods are based on 2 use cases - traffic counting using Nvidia DetectNet and pedestrian face mask detection using MobileNetV2. We also discuss how video frame resizing can be used for enhanced processing on the edge device. Index Terms— Edge Computing, Machine Learning, Smart Cities. I. INTRODUCTION According to Wikipedia, “A smart city is an urban area that uses different types of electronic methods and sensors to collect data. Insights gained from that data are used to manage assets, resources, and services efficiently; in return, that data is used to improve the operations across the city [1]. Smart cities usually have a lot of traffic camera deployments throughout the city. The footage from these traffic cameras is used to gather insights for better planning of smart cities. However, there are 3 major issues to video analytics performed at the centralized server. They are Network Bandwidth, latency in real-time situations, and citizen data privacy [2]. Processing the videos at the centralized server requires huge bandwidth to transfer the feeds to the cloud and is in fact not real-time which limits the benefits of the analysis. Moreover, In today’s connected world, people are more and more concerned about their privacy and do not prefer the idea of centralized analysis. Edge computing is proposed as an alternative to traditional cloud-based processing which helps in resolving the above challenges. Edge Computing is a new distributed Cloud Computing paradigm that describes a decentralized way of computing and storage at the topological edge of the network, i.e. in physical proximity to the data sourcing device. In contrast to the Cloud Computing paradigm, where the collected data is fully transmitted to a central server before being analyzed and used, Edge Computing allows to do these tasks on devices or close to it [4]. Grenze ID: 01.GIJET.8.1.15 © Grenze Scientific Society, 2022 Grenze International Journal of Engineering and Technology, Jan Issue