Object Detection with Deep Learning for the Visually Impaired

Students: Andre Settles, Prithvi Gali, Cheima Aouati
Advisor: Dr. Lin-Ching Chang

Visual impairment is one of the most prominent disabilities in America. According to the CDC, about 4.6% of the general population have some form of visual impairment. The visually impaired often rely on aid to navigate, but some existing aid (such as the white cane or service dog) can be limited in functionality. A possible innovation for visual aid would be to utilize deep learning and a wearable camera to detect and identify objects in a public environment. Object detection is a technique in which a software program is able to locate instances of objects within a digital image or a video and then determine the class of the objects. With the use of deep learning algorithms, we are able to train a model that can accurately detect our classes of seven different objects. With our model trained to an acceptable level of accuracy, we deployed the model on an Android phone, which is able to detect objects in real time with its camera. We then incorporate text to speech technology to audibly inform the user of exactly the objects that are in front of them.

Building Nutri: A Feature-Rich Personalized Diet and Fitness App

Students: Christian Attorri, Geetanjali Sharma, Michael Quinn
Advisor: Dr. Lin-Ching Chang

It’s no secret that America is in the midst of a growing obesity epidemic, as unhealthy diets and sedentary lifestyles continue to remain prevalent in the country. Many popular diet and fitness mobile apps have appeared to aid users in losing weight in order to counter this dangerous trend, yet they often lack key functions and require expensive membership fees. Our app, Nutri, seeks to improve on what is currently on the market while remaining entirely free. Nutri offers personalized diet and workout plans based on the user's physical condition and specific dietary requirements. Users can also track other foods they eat using the integrated barcode scanner and search functions. In addition, we want to use progress tracking and gamification. The progress tracking will show the user how close they are to hitting their goal weight, motivating the user, and gamification will be employed to make the user experience enjoyable. On top of these features, Nutri utilizes machine learning to provide further dietary guidance based on one’s diet history. After finalizing our application and its functions, we plan to deploy Nutri on both the iOS App Store and Google Play Store.

Unending Garden: A Video Game with Adaptive AI

Students: David Olumilua, Sebastian Scheller, Thea-Jeanne Trinh
Advisor: Dr. Lin-Ching Chang

Video games have gained immense popularity and offer entertainment for all with their diverse genres, companies, and communities. While AI is a crucial aspect of gameplay design, it has not advanced as much as graphics and visuals. Most games have a set difficulty level for battling AI, which does not adapt to the player. In this project, we aim to develop a 2D game with adaptive AI; AI-controlled enemies would learn the player's tactics and movements to find better and efficient ways to kill them. Using adaptive AI can enhance player experience by avoiding repetitive enemy behaviors and adding difficulty and immersion. By incorporating machine learning algorithms for the AI and C# to program the gameplay, we plan to create an endless game that adapts to the player's actions and play styles.

Machine Learning for Trash Detection in Waterways

Students: Wesley Garnes, Rhea Roxy, Ellen O’Brien, Blaise Trapani
Advisor(s): Dr. Rebecca Kiriazes, Dr. Jason Davison, Grace Pooley Deans, Dr. Matthew Jacobs

Waterways in urban areas are susceptible to high levels of trash pollution. In addition to environmental concerns, maintaining a high standard of water quality is essential for the preservation of waterways for recreation, agriculture, and industry. Our project, the Anacostia Waste Detection Project (AWDP), will make a water-based object detection machine learning model available to the public. The project team's immediate goal is to create machine learning models that can accurately detect surface-level trash in waterways. The autonomous waste detection system will generate accurate statistics useful for targeting efforts at water clean-ups, locating sources of pollution, and informing local policymakers. The project team developed the first iteration of the AWDP using the Trash Annotations in Context (TACO) dataset's Mask R-CNN model for image detection. This first iteration struggled to accurately detect and classify images of trash in a water environment. We are developing additional image detection models using a custom dataset. The team curated this dataset using the image annotation program Label Studio and video collected from the Anacostia River in Washington, D.C. We will expand the dataset to incorporate additional locations and publicly available data. The current AWDP prototype requires familiarity with the programming language Python which presents a significant barrier of entry for users. The project team will equip AWDP with a graphic user interface (GUI) designed to accommodate the greatest range of users by simplifying the experience while still providing flexible features. In making the AWDP publicly available and accessible, we aim to create a tool for managing trash pollution in waterways.

AOC Lot 7 Redesign

Students: Vincente Johnson, Connor Quinn, Victoria Roscoe, Samantha Scian
Advisor: Dr. Rebecca Kiriazes

DUALSx2 will redesign a underutilized parking lot, Lot 7, near the Rayburn House Office Building, into a park and maintenance facility to meet the Architect of the Capitol’s needs and address current site challenges. The site is located in the 50 and 100 year flood plains and experiences water pooling on the southern half of the site. A combination of bioswales and rain gardens, informed by calculations from OpenHydroqual, will help combat the issue of flooding and water flow throughout the site. Our proposed rain garden system features native plantings chosen for their hardiness, water absorption and flooding mitigation capabilities, low maintenance, and cultural significance. The site will also include a new shop building for the grounds crew of the AOC, with a minor structural analysis performed for the proposed building. Walking paths (analyzed using Bentley LEGION) are adorned with educational signage explaining the history of the U.S. Capitol Stones and the park’s landscape system details. This project will help the AOC meet US Botanical Garden 2023 goals related to sustainability, education, and maintenance. Final deliverables will be presented
in a report to the AOC.

Grand Engine Turbo Outfitting

Students: Chase Dreitlein, Leonardo Giron Berrios, Malik Alghamdi, Faisal Alenezi
Advisor: Dr. Jandro Abot

The Grassroots Motorsports Challenge is an annual competition involving penny pinching ingenuity and speed. The objective involves teams buying and modifying a car for $2000 to compete in three major events including an autocross, a drag race, and ending with a carshow. The objective of the Grand Engine Turbo Outfitting Engineering team is to design, build, and retrofit a 1998 BMW 540i with a turbocharged system. The original design included two turbochargers, which would keep the system symmetrical on both sides of the engine to improve the airflow from the exhaust back to the turbochargers. The preliminary design was simplified to use a single midsize turbocharger and reuse the current exhaust mounting points. Then will be routed on the outside of the car to the engine bay. Using SimScale modeling the team determined the most efficient way to use the airflow from the exhaust system and reroute it to the single turbocharger. Both sides of the exhaust system are merged together and routed to the exhaust inlet of the turbocharger. The final configuration should provide an increase of 10 psi of boost at 3500 rpm.

Hydraulic Flume Weir Testing

Students: Nate Perrins, Meg Metzger
Advisor: Dr. Rebecca Kiriazes

Hydraulics is a technology and applied science using engineering, chemistry, and other sciences involving the mechanical properties and use of liquids. At Catholic University hydraulics is a required class for Civil and Environmental engineering students to learn the properties and uses of liquids. The hydraulics lab contains a hydraulic bench which is used to manipulate flow rates and test different mechanical and hydraulic processes. It’s also equipped with a hydraulic flume to demonstrate flow in open and wider channels. The flume was unconnected to the hydraulic bench and thus unusable. Our team used plastic tubing to connect the hydraulic bench to the flume. Once we connected the hydraulic bench to the flume we added 5 different interchangeable weirs. A weir is a dam that controls the upstream water level. The weirs were designed in OnShape and printed with a 3-D printer. We tested our complete design in the hydraulics lab. Lastly, we determined what instrumentation is needed to use this structure for the education of students on topics such as weirs and hydraulic jumps. Overall, we built a hydraulic flume system with interchangeable weirs for educating future students.

Sustainability at the Federal Reserve

Students: Riley Meyers, Mohammad Alajmi, Bradford Wood
Advisor: Dr. Rebecca Kiriazes

The Federal Reserve Board in Washington, D.C. is located on Constitution Avenue directly across from the National Mall. The 1951 building was built in 1933, the Eccles building in 1937, and the Martin building in 1974, as such, the LEED program did not yet exist. The Martin building recently underwent a substantial renovation to include LEED Gold components, however the other two buildings remain LEED deficient. In 2017, Washington, D.C. was named the first LEED Platinum certified city in the world. The 1951 building, which contains the mechanical room for all of the Federal Reserve Board buildings, is now going under major renovations and new construction to become LEED certified. To accomplish this goal, WCC has begun a preliminary design process for the Federal Reserve. LEED requirements state the building must register 80 points or higher to be certified at a Platinum level. WCC will achieve this goal by providing recommendations based on the LEED v4 manual. These recommendations will come from the analysis of seven LEED categories and their subparts to determine which components will best suit the building based upon site conditions.

Student Steel Bridge Competition

Students: Mitchell Aaron, Megan Kuhns, Emily Moriarty, Nathaniel Selden
Advisor: Dr. Jason Davison

Each year ASCE and AISC hosts a Student Steel Bridge Competition that challenges students in a hands-on steel-design project. The objective of the Catholic University team is to design, fabricate, and construct a scale-model steel bridge that spans 23.5 feet and can carry 2,500 pounds, according to the competition rules. The problem statement is based on the Sweetwater River Bridge at the Wildlife Refuge in San Diego. The judging criteria of the bridge includes aesthetics, construction speed, lightness, stiffness, construction economy, structural efficiency, and cost estimation. The steel model bridge is required to fit a construction zone that will mimic the river environment. The model consists of multiple members with bolted connections. Our design was created based on structural efficiency, weight reduction, and constructability. The final design of our bridge consists of one row of seven stringers and sub-stringers for additional support. The main judging criteria is based on the overall deflections that test the structural durability of the bridge itself. The vertical load test will place 2500 pounds and the lateral load test will consist of 50 pounds.