Smart object counter

• Main Author: Ting, Rachel Jie Yi
• Other Authors: Loke Yuan Ren
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/137919
• Institution: Nanyang Technological University

Crowd counting has been a prevalent problem in the field of machine learning. There have been numerous researches done to improve upon the accuracy of existing solutions. This project aims to incorporate Crowd counting into the context of Universities to provide students with a more accurate estimation of the number of people at various locations around the campus. This would allow users to make better informed choices and plans on when and where to head to. For this project, the images are taken from the NTU website which displays the CCTV images of several locations around the NTU campus. Deep Learning libraries in python are used to estimate the number of people in those images and make a forecast of the crowd count in the future. A few examples of the libraries that will be used in this project would be Pytorch and Theano. The required data would then be stored in a Firebase Realtime database to allow the web application to retrieve the data easily and efficiently. For the front end, HTML programming is used to design the user interface, whereas JavaScript (JS) is used to control the behaviour of the web application. After analysing the results, the crowd estimation for the various locations are largely accurate. However, the estimation at a few of the locations during night-time does not correspond with the expected count. Forecasting proved to be fairly accurate and is able to display the expected crowd behaviour throughout the day, though the predicted count might differ from actual. Overall, the objectives of the project were met with considerable success. However, there are still further improvements that can be made to increase the accuracy and functionality of the application.