Moving Objects Counting Dashboard Web Application Design

The ability to detect and count people using digital image processing techniques has a lot of applications. Retail stores need to understand foot traffic to organise their goods, understand peak times or prevent theft. Business or government agencies could use the counter to design or manage buildings and other public places. A system of object detection algorithm runs on CCTV footages needs a good platform to display the result that easy to access and interactive. This paper will explain the design of dashboard web application used to display the number of objects in a CCTV video and in our implementation the objects are people in a building and vehicles on the traffic intersections. The main focus of this paper is the novel software architecture that is consisted of four different programs that are communicating with each other. Those programs are publisher, object detection, video combiner, and dashboard program. Object detection system is using YOLOv3, an algorithm using deep learning and convolutional neural networks (CNN) that can detect objects fast and real-time. The terminology "real-time" means that the web application can detect the number of objects in a live CCTV video with few seconds of delay. The computer specification that is used in the test to achieve real-time result is Intel Xeon E3 CPU, NVIDIA GTX1060 GPU, and 8 GB of RAM. OpenCV, an open source library for computer vision and machine learning is also used. The dashboard design is using Dash by Plotly, an open source Python based framework to create web application.