This project explores the effectiveness of three machine learning approaches—Poisson regression, U-Net, and Scale Aggregation Network (SANet)—in estimating crowd sizes across different environments. Using two datasets, including the ShanghaiTech dataset and a mall surveillance dataset, we assessed model performance in varying crowd densities and conditions. Our findings indicate that while Poisson regression is computationally efficient, deep learning methods like U-Net and SANet provide significantly higher accuracy. SANet, in particular, excelled in dense crowds but required longer training times. This study contributes valuable insights into the trade-offs between accuracy and efficiency in crowd counting models, highlighting their potential applications in public safety and event management.
This project was done in a group of 3 people. My main contribution and ML method was SANet Methodology & Evaluation
Written with Python
HitsOnDeck is a web application built using HTML, CSS, and React, designed to simplify the process of buying and selling tickets for music festivals and concerts. The platform offers users a seamless experience to browse, book, and manage event tickets, as well as create and sell their own festival events. Key features include a booking history dashboard, an event creation tool, an interactive ticket purchasing system, and a personalized event recommendation section. The design prioritizes user-friendliness, with a modern UI and intuitive navigation, making it an ideal solution for festival-goers and event organizers alike.
2024
Java Application
This desktop application, developed in JavaScript, is designed to enhance productivity by blocking access to websites and applications for a user-defined period. The system supports multiple user accounts, allowing each user to configure custom blocking presets based on their study or work needs. A key feature of the application is its blocking screen, which enforces focus by preventing access to restricted content while the timer is active.
My primary contribution to this project was implementing the blocking functionality and screen overlay, ensuring seamless enforcement of restrictions for an optimized distraction-free environment.
Flynn Kassulke 