CAPSTONE DESIGN PARTICIPANTS

Yifan Li

Yi Chen

Zhiming Lin

Shuhong Liu
Kraeq
36
This augmented reality device is intended to assist persons who are deaf or hard of hearing. The device's microphone features a voice recognition mechanism that converts speech to text. The text is projected in front of the users' eyes using augmented reality technology. Therefore, when chatting to others or listening to a speech, people with hearing loss can see the words. Furthermore, the AR device is detachable, so it may be worn with most glasses.

Zipian Zhou

Zhan Yuan Lu

Yifan Wang

Yaxuan Li

Shiji Liu
Expense AI
37
With the fast-paced living styles today, people forget an old tradition to keep track of their expenses. Meanwhile, they are increasingly surprised by the amount of money they spend when the monthly credit card bill arrives. In most cases, people are either too busy to book their expenses or are reluctant to do so for its tedium. The objective of this project is to design an all-in-the-one application that allows users to keep track of their expenses with minimal effort. ExpenseAI is a bookkeeping application that employs a variety of machine learning technologies so that users interact with the application as simply as taking a photograph of their receipts or making a few taps on the phone screen and asking ExpenseAI to do the rest of the magic. It uses the state-of-the-art optical character recognition engine to read the image, applies a well-trained classification model on top of the data and categorizes them automatically. ExpenseAI also allows the users to customize their own categories, export monthly expense reports and plan for a saving sprint. This design is favourable over the alternatives in the market since it expects least input effort from users but gives most output in return.

Michael Meng

Matthew Tang

Ken Ming Lee

Ethan Lee

Cliff Li
REALM-AI
38
As part of the game development cycle, game designers undergo a time-consuming and resource-intensive process of iteratively playtesting, re-designing, and fine-tuning their game before it is released. Using reinforcement learning, REALM_AI eases this process by training bots to play games, during which, gameplay data are generated, recorded, and analyzed by the tool. Game designers can then view the analyzed results to make data-driven decisions, without the need of a large playtesting team or other post-release analytic tools to provide these insights.

Andrew Xing

Aaron Tan

Clement Liaw

Ganashsai Vannithamby

Murtaza Fatakdawala
LiTTY
39
LiTTY is a deep reinforcement-learning based framework that aims to solve the challenging issue of playing highly complex games where the rules may be tweaked days before a tournament. Computers have dominated top human players in traditional games such as chess and Go; however, their rules have not changed substantially in hundreds of years. LiTTY produces a competitive opponent for Terminal Live in a matter of days — despite arbitrary changes to game parameters — by using self-play and experience gained while playing under a previous ruleset.

Andrea Martinez-Chung, Derek Cao, Jacob Cupryn, Mike Matyszczuk, Riley Thomas, Tiger Hao
Wally: Nursing Home Service Robot
40
Wally is a semi-autonomous service robot intended to be used in nursing homes. Its main purpose is to deliver water bottles to seniors, but it can be used for general delivery. An accompanying mobile app helps nurses set a desired path with checkpoints at pickup locations. The articulating arm on the mobile base allows for an arm movement to be configured by the user to manipulate an object at each checkpoint. The user can set a scheduled time for Wally to run a desired path.

Zhiwei Yan

Noah Metz

Faraz Sahba

Amir Soliman

Robbie Dubrick
iCoach - Personal Tennis Trainer
41
The iCoach is intended to be a closed loop tool to be used by Tennis players to monitor and analyze their game. The tool uses both hardware to measure a player’s performance, and software to process and synthesize the information for all players aiming to improve their quality of play.

Michael Fan

Illya Myshakov

Grace Tian

Daniel Zhao
Gautocos
42
Gautocos is an automated compost sorter that uses a machine learning model to separate compost and non-compost materials. Items are brought along a conveyer belt, scanned, and sorted accordingly. With Gautocos, the workplace with be less hazardous and more efficient for waste management companies.

Rushil Nagarsheth

Anton Vassilev

Anish Khanna

Alexandru Barbulescu
MyCircle
43
Over recent years many social-media users have felt disconnected with various platforms due to an increase in censorship and personal data violations, in addition to no longer feeling connected with their immediate circle. MyCircle is a platform that removes a centralized entity who can impose its influence on its users, while maintaining the security of users’ data in doing so. It also encourages tight-knit relationships by limiting the number of connections a user can make, keeping posts and interactions within smaller, more genuine communities.