Time series forecasting and imputation is an important area of machine learning. Our project explores the use of convolutional style deep-learning methods to solve time series problems. Implementing these systems involves comparing accuracy metrics between other state of the art methods. Our team worked on developing and testing models for performing forecasting and imputation on air-quality specific data. We hope that by solving these problems we can more accurately predict and impute time-series data.

Team Members: 

Rohan Balaji

Yuan Chang

Our project aims to provide an interactive experience that imparts a sense of community to its users. Additionally, it can serve as a prop to promote the Cockrell School of Engineering to potential students, donors and incoming freshmen. It is an interactive art installation in which a swarm of realistic, animated bees form the outlines of the user. The bees are attracted to the edges of objects that appear in the foreground of the video feed; when objects in the video feed move, the bees break away from their configuration to match the new edges in the video.

Oftentimes, images can be compressed without a human perceiving a difference. In our project, we train a deep learning model to replicate human perception of image quality based on data collected from a massive crowdsourcing study. This model could automatically predict the best compression level of images, potentially saving huge amounts of memory and bandwidth.

Team Members: 

Arkan Abuyazid

Shehryar Ahmed

Brandon Lee

Akash Shukla

Meenakshi Swaminathan

Kan Vanthanasuksan

Update and improve the existing ResearchMatch framework by fixing bugs, removing technical debt, and implementing new features for widespread use by the spring of 2020.

Team Members: 

Prajakta Joshi

Kenneth Kobayashi

Haosong Li

Tejus Mathew

Christopher Reedy

Samir Riad

Over the past semester, our team worked to update PrivacyCheck, a browser extension that uses machine learning to analyze any website’s privacy policy and then provide the user with easy-to-understand scores. As part of the update, our team reworked the entire user interface to make the design more polished and look more like a smartphone application.

nuCoach is a mobile application integrated with a pressure-sensitive mat that allows users to monitor their squatting form. Using computer vision, nuCoach can visually analyze your shoulder-hip-knee and hip-knee-ankle angles and your center of balance. The pressure-sensitive mat captures your weight distribution across your feet as a heatmap so you can see if you’re leaning too far forward or backward. Your reps are analyzed in real-time and are available for review through the mobile application.

Team Members: 

Haisun Banh

Allison Fang