Senior Projects Showcase

Spring 2024 - Computer Science Senior Project
"Look up and aim high" - Congratulations CS '24!

Natalie Lee, Carrie Wasieloski, Austin Counterman


The Riverscape video game features a classic pixel art style and a relaxing yet immersive gameplay experience for its users. Riverscape is an homage to the keystone species, beavers, and their important role in shaping river ecosystems' landscapes. By naturally modifying their surroundings, beavers transform rivers into biodiverse wetlands, essential for a healthy ecosystem. In Riverscape, players step into the role of a beaver. Thus, like a beaver, their primary objectives center on harvesting trees, building dams, and facing predators. As players progress and construct more sophisticated dams, they actively participate in transforming their environment. The game’s dynamic maps visually represent the evolving health of these ecosystems, illustrating beavers’ impact on their homes. Players complete the game when they successfully upgrade their dam to its fullest capacity, achieving the ultimate goal of environmental transformation.

Tech stack: Godot Engine, GDScript, Aseprite, GarageBand

Project Advisor(s):    Prof. G. Dimitoglou

Elijah Olsson, Subhashree Susindran, Jake Wantz 

“Park-It!”: Computer Vision Parking Space Monitor

Finding a parking space can sometimes prove to be a difficult task. The ”Park-It!” Parking Space Monitoring System introduces a computer vision-based solution to optimize parking management anywhere. Utilizing Linux, Python, and shell scripts with OpenCV and YOLOv5 for image processing, the system monitors real-time occupancy of a parking space through a USB webcam. Designed to enhance parking efficiency, it features occupancy detection, object recognition, and analytics to determine optimal parking times
and durations. Despite challenges such as data inconsistencies and transitioning from a remote to a locally hosted server, the team developed a functional minimum viable product, documented by over 1,500 lines of code and tracked all software development on GitHub. This project exemplifies the culmination of a computer science education, applying advanced computer vision techniques and analytics skills to improve urban infrastructure management.

Tech stack: Linux Python Apache PHP

Project Advisor(s): Prof. G. Dimitoglou

Awwal Ahmed, Sarah Freidel

ALEP: Advanced Loan Eligibility Predictor

Advanced Loan Eligibility Predictor(ALEP) is a sophisticated tool designed to refine the loan approval process. Developed using a Random Forest algorithm, ALEP evaluates potential borrowers based on factors such as income, employment status, and credit history. This automated system predicts the likelihood of applicants fulfilling their loan obligations by learning from historical data, ensuring decisions are both accurate and fair. By efficiently processing complex data, ALEP not only accelerates decision-making but also enhances transparency in loan approvals. Ideal for financial institutions aiming to modernize their operations, ALEP provides a reliable, user-friendly solution that simplifies and streamlines the lending process.

Tech stack: React, Flask, SQLite

Project Advisor: Prof. G. Dimitoglou

Corey Cevallos, Emanuel Granados


Econest aims to provide easy-to-use financial management to users looking for a solution to their financial challenges by combining income, expenses, and stock tracking into one intuitive application. We aim for users to effortlessly input, or upload bank statements, into our web application to gain valuable insights into their spending habits through the user-friendly interface. The Econest system operates on the modern MERN stack, leveraging MongoDB for flexible data storage, Express.js for seamless server-side logic, React for dynamic front-end components, and Node.js for efficient server-side processing. Integration with the Real-Time Finance Data API enriches the user experience by providing up-to-date stock prices on stocks they take interest in. We designed Econest for ease of use, allowing users to navigate effortlessly through their financial data. With interactive graphs and visual representations, Econest empowers users to gain insightful insights into their financial trends, and investment performance at a simple glance, enhancing their decision-making process. Econest stands as an asset for individuals seeking a straightforward and effective means of managing their financial journey.

Tech stack: MERN (MongoDB, Express, React, Node.js)

Project Advisor(s): Prof. G. Dimitoglou

Hood CSIT Spring 2023 Seniors


Christina Kong, Grayson Swift, Jaken Whipp

Computer Science Education Game

This project aims to provide students with a resource to practice their general computer science knowledge while also engaging in entertaining activities such as exploring an open-world environment and solving challenging logic puzzles. The objective of this project is to garner accessibility to Computer Science education. Our game provides an outlet for applying knowledge and demonstrates the significance of engagement in studying Computer Science. The expectation for this project is to ultimately engage the audience and promote Computer Science education.

Project Advisor(s): Prof. G. Dimitoglou

Humza Ahmad, Chriss Oboa

Travel Nest

Travel-Nest is designed to connect hosts and travelers, similar to the concept of couch surfing. Hosts can offer their rooms or spaces for travelers to stay in, either for free or for a small fee. The rooms and spaces are displayed on a map or in the form of cards on the page, complete with detailed information and a slideshow of images. Travelers can easily filter the available rooms by city or cost, making it easier to find the perfect place to stay that fits their budget and location preferences.

Project Advisor: Prof. G. Dimitoglou

Julian Urban, Brian Hitchcock

Cybersecurity Exploration

Game Cybersecurity is a rapidly growing field of recognition and community for those wishing to enter the area. As such, there have been multiple strategies to assist with teaching cybersecurity to new learners. This Project was done to help teach computer science students about cybersecurity, all while doing so in an environment that allows the learning process to be done in game form. The game was developed using Unity as a game engine and editor while utilizing C# for the backend development. While the effectiveness of this has yet to be proven, background research suggests that teaching students in an entertaining manner has been successful when it comes to general comprehension of the material.

Project Advisor: Prof. G. Dimitoglou

Jack Carr, Richard Williams

Adventure Around The World, Over Night Express

Adventure Around the World is a multimedia, single-player game written in Python using the Pygame libraries. The game allows players to experience different locations and sights worldwide through a series of playthroughs.

Project Advisor: Prof. G. Dimitoglou

Aalayah Honablue, Salman Alrashidi

Good Eats

"GoodEats" a cross-platform mobile application developed using React Native and Firebase is restaurant goers dream! The app aims to provide food lovers a platform to share their opinions about restaurants and discover new dining experiences. GoodEats combines Yelp and Twitter, where users can post reviews and ratings of their favorite (or not-so-favorite) eateries. The app allows users to search for restaurants based on cuisine type and location. Users can follow other foodies and see their recommendations in a personalized feed. The app also includes a bookmark feature that allows users to save restaurants they want to try in the future. GoodEats is a user-friendly, engaging mobile app perfect for foodies who want to explore new culinary delights and share their experiences with others.

Project Advisor: Prof. G. Dimitoglou

Brandon Knotek, Walid Muhammad, Jack Wilder
"Mind Games"
Faculty Advisor: Prof. G. Dimitoglou

Computer systems are generally not designed to be accessible for people with severe physical disabilities. Brain-computer interfaces (BCIs) present a potential solution to this accessibility problem by eliminating the need for physical interaction with a computer. Our study was focused on exploring the viability of a consumer-grade BCI for real-time interaction with computer software. To this end, we developed a simple game of Pac-Man and supporting software that receives EEG signals and classifies them into commands to be executed in-game. Our results were promising and indicated that a consumer-grade BCI could enable non-physical interaction with computer systems. However, overall command classification accuracy was relatively poor. Further experimentation and fine-tuning are required before our system can be useful to the public.

Tech Stack: Open BCI, Python, Flask

Andrii Bezmen, Jesus Lopez
"Fighting Diabetes"
Faculty Advisor: Prof. D. Sierra-Sosa, Prof. G. Dimitoglou

Scientific research into Diabetes is a domain rife with large amounts of data waiting to be extracted for the purpose of identifying and disseminating meaningful healthcare information. This project aimed to apply clustering techniques, such as K-means and spectral clustering, to a private healthcare dataset from Spain to identify subgroups of tightly related comorbidities among patients with Type 2 diabetes. In this project, we used clustering techniques inside the popular Scikit-Learn machine-learning library to investigate which, if any, subgroups of comorbidities would be revealed and yield meaningful analytic insights. By the end of the research period, while their subgroups were identified and consistent results were achieved, explaining the results to a satisfactory standard was not possible with clustering alone, leading to an attempt to use other techniques such as formal concept analysis and decision trees to create better explanatory models. Given the inconclusive results of those other techniques, further work is needed to explain whether the clustering yields any actionable identification of risk factors for more severe symptoms of Type 2 diabetes.

Tech Stack: Python, Conda, Arize, Scikit, Numpy, matlibplot

Alex Algazi, Diana Teka*
"Discord SET Bot"
Faculty Advisor: Prof. G. Dimitoglou

There is no reliable way to play SET, the pattern recognition card game, online with friends. In the modern internet age, a game’s ability to survive and stay relevant depends on its accessibility in the online space, so we set out to make a Discord bot that would facilitate multi-user games of SET. The bot must allow all users to interact with the same board and track player scores in a leaderboard. This functionality was achieved by using multiple event listeners, one for each row of cards on the board, registering user inputs, and caching game data. In the end, we now have a bot capable of hosting games for many users in a Discord server and robust stat tracking capabilities. Now, in 2022, we can finally play SET online with an easy and free solution, with the only caveat being that each user must have a Discord account.

Tech Stack: JavaScript, Node.js, Discord.js, SQL Lite

* Deceased (March 2022)

Gaganpreet Kaur, Veronica Stewart, Nathaniel Rowe 
"Money Opportunity Application"
Faculty Advisor: Prof. G. Dimitoglou

The Money Opportunity application help users meet their financial goals by providing suggestions and financial analytics to make informed decisions. This application collects user financial information, displays analytical information, and recommends how to save money. To make suggestions, this application collects user information such as current job titles, skills, income from all sources, expenses of all types, and expected retirement age. The application provides various financial visuals to learn about project income, savings, and expenses. Also, the application provides job recommendations based on the user's job title and skills. In addition, the web application displays live stock data of top companies and relevant news based on users' job titles. Moreover, users can search for various jobs within this portal. Users can also use the retirement calculator to better understand how much money they will have during retirement.

Tech Stack: Laravel, PHP, MySQL, Bootstrap, Docker

Seth Jones, Shelby Vaughn
Faculty Advisor: Prof. G. Dimitoglou

Creation of a simple and accessible website that can allow for the conversion of text into speech and microphone inputs into parsed text. The project includes options for voice-altered rendering and mascot character/avatar association with different voices.

Tech Stack: HTML/CSS, JavaScript, IBM Text-to-speech API, Google’s Speech-to-text API

Oluwatobi Aroloye (Tobi), Sorupa Wagle, Christopher Smalls 
"Stock Web Application"
Faculty Advisor: Prof. G. Dimitoglou

To an individual with prior knowledge of the stock market and how the overall process works, this program is self-explanatory. However, this application allows users to gain an easy and reliable introduction to stock portfolios, thus, allowing them to have familiarity with potential real-life transactions in the future. The project is an application developed to mimic the process of investing money into a stock and monitor its changes over time through included features. The simulation achieves its purpose of being a straightforward program by allowing an individual to perform the abovementioned actions.

Tech Stack: Python, Django, JavaScript, ReactJS, SQLite

Kyle Hinton, Paul Wells, Yohannes Terefe, Somayyeh Kamyab, Kyle McQuillen
"Creating a First Person Shooter (FPS) AI through Deep Learning in Unreal Engine"
Faculty Advisor: Prof. G. Dimitoglou

The video game industry is a highly profitable market in the world tonight. The market and scope of individuals involved in video games will only continue to grow as we move further and further into the technological age. As many gamers are drawn more and more to multiplayer interactions, there has been less focus on single-player games. Many can argue that as they develop, the computer characters in-game are too easy to play against. This project proposes and explores a way of creating a more human-like computer character that can challenge players in a single-player environment. A new computer character can be created through a neural network and reinforcement learning, which closely reflects a real-time player's actions. The model demonstrated impressive cognitive ability and development through the additional training it received in testing. With more testing in the future, the hopes of this AI model being placed in solo RPGs worldwide could revolutionize the enemy computer for years to come.

Alex Blank, Juan Pablo Chacon, Natalie Hani, Lauren Thompson, Jonathan Wallace
"An In-Depth Analysis with Visual Representation of the US Criminal Justice System"
Faculty Advisor: Prof. G. Dimitoglou

Previous research has suggested that there has been overpopulation and overcrowding of prisons. We can see that this impacts our society and goes deeper down to an individual level. The problem we wanted our team to solve was to find a way to develop analytical data to explore the prison micro-systems and crime data through the use of data sets we have procured. We wanted to go through various ideas like prison demographics, gender disparity, and prison population through various states and prisons inside the United States. Our solution was to make a dashboard to visualize the data we have obtained through the data sets that enable dynamic interaction and highlight issues within the prison system. The results were that we were able to make various comparisons between states, gender sentences, and other demographic disparities. For the dashboard itself, I believe it was a good way to represent all the data we have found and were able to show through the use of the d3.js JavaScript library.

Trenton Lazorchak, Olanrewaju Aribisala, Jake Howell, Rediat Uregessa
"Booze Buddy" (A Mobile, Geo-aware  Comparative Liquor Store App)"
Faculty Advisor: Prof. G. Dimitoglou

Now that mobile use has increased creating an application that allows users to get information faster and easier is more important than ever. Mobile application users are always looking for the best way to find the information they need, specifically for finding store prices. The process now is obnoxiously long and can be simplified with the creation of an application. To save people time, the app Booze Buddy was created. Booze Buddy takes in your current location and provides all the liquor stores in your area. It also shows the drinks in each store, allows the user to easily compare prices, and saves time looking into each store's catalog. To save the user time, this mobile application was created to provide the user with the easiest access to price and drink information, store information, and ratings.

Michael Erik Barnhart, Robert Atkins
"The Evaluation of Machine Learning Algorithms Through the Classification of Image Data"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: Given the advent of machine learning and the rise in popularity of the approaches the discipline offers to solve problems, we thought it prudent to investigate whether or not a certain algorithm would perform more efficiently than another when categorizing image data. Given that machine learning algorithms can be applied to a variety of fields, such as analyzing financial data, looking for trends in power usage across an organization, or identifying patterns in images. Identifying anomalies in biological systems is an important indicator that something could be wrong with the body. Because the human body consists of trillions of cells, it can be hard to identify cells that are exhibiting abnormalities, and thus can be overlooked by the human eye when reviewing for a potential problem. To rectify this, doctors have turned to machine learning to help them identify abnormal cells. Because there are a number of machine learning algorithms that can be implemented to perform image identification, we wanted to test three different algorithms to see if any provided an efficiency advantage over the others. We elected to use a Convolutional Neural Network, an Artificial Neural Network in the form of a Residual Neural Network, and a Support Vector Machine. We chose to have each system analyze malarial cells. We found that the Artificial Neural Network was the most efficient of the three models, followed by the Convolutional Neural Network, then the Support Vector Machine. The most accurate model was the Convolutional Neural Network with an average of 86%, with the worse being the Support Vector Machine and Artificial Neural Network at around 63% for both. The longest to train was the Convolutional Neural Network at an average of 900 secs per run, followed by the Support Vector Machine at 700 secs per run, and the fastest is the Artificial Neural Network at just over 200 secs. However, we did find that the loss for the Artificial Neural Network approached near zero very quickly, resulting in a learning plateau that we believe would not be able to overcome with additional training resulting in a model that would not improve past 70% accuracy.

Michael Brady, Matthew Dees, William Edge
"Monumental Anxiety"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The objective of this project was to develop an interactive mapping web application that allowed users to explore Washington DC monuments. Using a service like Google maps to figure out the locations of monuments is a tried and true solution, but we wanted to build a product that would expand upon this and provide the user with a more in-depth exploration of different locations, along with information about each particular monument. We approached this problem by creating a full-stack web application utilizing HTML, CSS, JavaScript, PHP, MySQL, and an interactive web map API named MapBox. Our result was a front-end user interface connected with back-end app logic to store/display information from a file system and database. It also included an administrator front-end user interface that allows our client for this project to delete, edit, or input information directly into the web application. We as a team feel we have created a project that incorporates all of the requirements and could serve users as an interactive way to learn about different monuments in our nation’s capital.

Joshua Ly Soumphont, Danielle Elise Pitts, Jacob Edward Saintcross
"Student Shift"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: As students who are graduating and headed into a new chapter of our lives, we see and have to experience the lack of knowledge about jobs out there and possible options that we have. That is why we have decided to develop Student Shift. Student Shift is here to help bridge the information gap between students and life after graduation. This tool is a web-based data analytics and visualization platform that takes a populated database of post-graduate data analyzes it based on a query, and provides visuals of past students’ academic trajectory and employment.

Katherine Kladky, Kaitlin Marutani
K and K E-Voting
Faculty Advisor: Prof. G. Dimitoglou

Abstract: Online voting is already employed by various individuals and organizations, through various concerns have prevented it from more widespread, official use. The purpose of our project was to make a functional method of online voting that would be both easy and efficient to use and secure. We created an e-voting platform with a website front-end and a database back-end using CakePHP and MySQL. Our current project functions well and enables users to create surveys and recruit respondents. It also allows the end-user to easily vote and view the survey results. The K & K E-Voting platform works and is effective and efficient.

Alain Shekanino, Matt Blank, Dylan Blevins
"Roundabout vs Traffic Light Intersection Traffic Flow Optimization"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: Have you ever sat at a traffic light and thought to yourself "there has got to be a better way". In this paper we plan to provide solutions to this problem. We wanted to know which is more effective at redirecting traffic, a roundabout or a four-way intersection regulated by traffic lights? We approached our problem by analyzing average waiting times and vehicle queuing lengths. We found that a standard conversion from an intersection to a roundabout provides over 50 percent decrease in vehicle waiting times at the road junction, and a further 50 percent when the roundabout is further optimized. We suggest that newer road developments should incorporate roundabouts and current intersections should be converted into roundabouts if possible.

Rachel Shafer, Natalie Shafer
"Drift Simulator"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The objective of this project is to create an application that calculates and displays the most probable position of an object adrift at sea within the Gulf Stream area. This will be accomplished using ocean current and wind data to calculate the influence it has on an object. This project is centered around the creation of an ocean drift simulator based on ocean currents and weather conditions. The importance of this project is based on the fact that search software like this is imperative in order to prevent unnecessary loss. It could be used as an educational application to see the effects drift has on an object. This drift simulator would provide more accessible software that can be used to find objects lost at sea, to the public. Individuals who wish to search for things lost at sea could use this application coupled with data surrounding the time of the event, to locate objects and people. This could allow more individuals to help in these searches, which would result in a higher probability of survival for the survivors lost at sea. It would also be lucrative for businesses that would need to recover the lost cargo. The purpose of this project is to develop an accurate drift simulator that can be used to find the most probable location of an object lost at sea given the most probable last known position of the object, the time elapsed, and the specification of the object or its dimensions. The program will calculate the approximate final location.

SP2019.pngBrandon Ubiera
"Analyzing Hashtags Using Sentiment Analysis"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: Tweets with specific hashtags from Twitter have been collected using a Twitter developer account. The hashtags that were collected are Facebook, Amazon, Apple, Netflix, and Google. Each hashtag will be its own data set. Stock data has been collected that correspond to each hashtag for the corresponding dates. Sentiment analysis was performed on all of the Twitter data sets to determine if each tweet is positive, negative, or neutral. A database was built to hold all of the tweets and stock data. A website user interface (UI) was built to view the data from the database and allow users to choose different criteria, and the website UI will dynamically generate graphs. These graphs can then be used to analyze the data to find interesting results about Twitter, Twitter users, as well as companies, companies' stocks and items users tweet about.

Chris Wetzel, Shaun Sullivan
"Pedestrian Flow Dynamics"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: Pedestrians prefer to move in social interactive formations. We aim to determine which formation shape is the most efficient in reducing collision time and group dispersion while traveling a short distance. We have built a simulation in Unity3D, where the formation shape and speed of each pedestrian can be controlled to simulate group collisions. We measure group dispersion and time taken to travel, disperse and regroup to determine the most efficient group formation. Our results show a Concave V formation of three pedestrians displaces and regroups 10% faster than the slowest which is a side by side formation, while the side by side formation has a 30% lower group dispersion than the Convex V formation which has the highest. Our results apply to a narrow environment, such as a sidewalk, hallway, alleyway but can be generalized for other environments.

Mickayla Storm Bachar, Joseph Steven Carroll, Zachary Michael Crossey
"Roommate Matching Software"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The current roommate matching process at Hood College does not allow for considerations related to personality traits or preferences. This paper describes the Roommate Matching Software System (RMS), designed by Team Optimism at Hood College. The objectives were to create an intuitive and meaningful questionnaire to gather personality information about students, build a database to store and retrieve updated information, and use Irving’s Algorithm for the Stable Roommate Problem (SRP) to find a stable matching for the current students in the set. The stable matching would contain the ideal roommate for the student running the algorithm. Another major objective was to create a website to allow the students access to the system. Using this system as part of the roommate selection process should increase the likelihood that roommates at Hood College, with similar preferences, have the ability to find each other.


Joshua Clemens, Cory Watson
"A Visualization of Honeypot Activity"
Prof G. Dimitoglou:
Abstract: This project was developed to study the activity of malicious attacks on a variety of platforms. We accomplished this by creating a honeypot server, consisting of: a SSH honeypot, Kippo, a HTTP honeypot, Glastopf, and a Windows Server honeypot, Amun. Log parsing scripts were developed to extract meaningful information from the raw log files generated by our honeypot. We then created a web-application to use as a dashboard to visualization this information.

Karen J. Canas Hernandez, Jacob W. Denion, Afton J. Woodring
"Web-enabled Repository & Note-taker Scheduling System Revision"
Prof E. Chang:
Abstract: The system being developed is a web-based repository and note-taker scheduling system. Its goal is to replace an older, ineffective, manual process. The web-enabled database will provide an automated and more efficient process as it will allow the note takers to upload the documents where the students can easily access them. The system will only be usable by those that have been deemed as note-viewers, note-takers, and administrators. While this system is being designed for a specific instance, it could be generalized to fit the needs of other departments that have need of a note taking application with a database.


SrProj2017.pngRyan Dormitzer, Geoffrey Huntoon, Gary Lopez
"Face Off"

Abstract: Face Off’s project is an attempt to use open-source software to replicate an algorithm that detects irises and stores them as barcodes for later comparison to “employees.” The project would ideally be part of a cohesive security system. Such barcode transformations exist, but, according to our research, implementations are not openly available to the public. The project is coded in Python and uses OpenCV for face detection and iris extraction. PyQt is also used for graphical user interface and testing purposes.
The process of storing irises, in brief, is:

1. Use Haar cascades in OpenCV to detect a face in an image, and subsequently the eye region and individual eyes

2. Use OpenCV’s HoughCircles function to draw circles around the iris and pupil in order to extract the iris from the image

3. Apply feature value extraction and normalization using a Gabor transform and Ordinal measure, as described in Optimal Generation of Iris Codes for Iris Recognition.

4. Unwrap and binarize the normalized iris into a barcode-like object, again as described in Optimal Generation of Iris Codes for Iris Recognition.

5. The barcodes are then compared to eyes found in a live video feed.


Thomas Corcoran, Clark Spessard
"Predicting the Spread of Wildfires with Artificial Neural Networks"

Abstract: Wildfires are a significant threat to agriculture, wildlife, and human life and property. Climate change has caused a dramatic increase in the number of large-scale wildfires (> 100 acres) that occur in the United States each year, and the U.S. Forestry Service estimates that an area the size of Colorado is at risk of destruction from severe fires in 2017. For the first time in its history the Forestry Service spent over 50% of its budget fighting wildfires in 2015, and this number is expected to reach nearly 66% of its total budget by 2025. Because the act of fighting a wildfire is fundamentally a resource-allocation problem, and therefore an optimization problem, we determined that with sufficient data the problem could be tackled in real time via an integrated data mining and machine learning approach. Initially it appeared as though high-quality data were in sufficient supply: the U.S.G.S.’s GEOMAC system contains shapefiles of active wildfire perimeters updated on a daily basis stretching back over 15 years, and web services such as the Weather Underground provide coordinate-based weather data, including on a historical basis. Unfortunately, the vision of a real-time perimeter prediction system proved infeasible due to extreme irregularities with the GEOMAC data (missing values, inconsistent measurement techniques, non-standard formatting, etc.). The notion of an artificial neural network, trained on historical wildfire perimeter data, predicting the evolution of active wildfires based on their current coordinates and nearby weather and topological data was still irresistible, however, and so in an attempt to salvage the project we endeavored to generate our own data with which to train a network and provide a proof-of-concept for the idea instead. We generated our datasets by implementing a basic, but mathematically sound, wildfire spread model in Python. Using this model, we simulated many instances of wildfire growth that we then used to train a Keras-based neural network adhering a more-or-less standard multi-layer perceptron form. Our system is able to predict the growth of a wildfire’s perimeter with ~90% accuracy in our testing. Our final results tentatively indicate that the originally envisioned system is in fact quite realistic, once the issue of data sourcing is resolved.


Rafael Zamora, William Steele, Joshua Hidayat & Lauren An
"DeepDoom: Visually Navigating 3D Environments Using Distilled Hierarchical Deep Q-Networks"

Abstract: Deep Q-Networks (DQNs) heavily aided in guiding agents to associate differences between rewarding and penalizing actions on both 2D Atari 2600 games and 3D Doom games. To compensate more effectively for the dimensional complexity of a 3D environment such as Doom, Hierarchical Deep Q-Networks (h-DQNs) allowed us to modularize tasks within this environment into separate skill-based models—our basis for designing more efficient agents. Building upon previous research, we implemented h-DQNs within Doom to exploit the advantages of a hierarchical structure in integrating separately-skilled tasks into one network of knowledge; furthermore, we employed Policy Distillation onto h-DQNs to reduce the number of resources needed while retaining high-level knowledge. Our trained h-DQN model provided positive indication of the agent’s ability to combine vastly differing skills, and saw both an enhancement in its skill-adaptive capabilities and training times when compared to a single DQN. Through Policy Distillation, we also transferred knowledge acquired from our h-DQN into a single DQN architecture, allowing for an overall reduction in computations and network parameters. This led us to measure the effectiveness of distilling a pre-trained h-DQN in its ability to retain key information from the high-level knowledge and lessen resource-intensive computations.


William Bell, Rafael Torres, Michael Yura
"Utilizing OpenCV to Improve Athlete Performance in Pole Vaulting"

Abstract: The pole vault has been an Olympic sport requiring that an athlete display a masterful combination of speed, strength and technique for over a century. For many athletes, competing in the Olympics is an honorable, patriotic and challenging task where victory translates into global prestige and towering achievement. Years of grueling labor, emotional tribulations and constant proper repetition allow an athlete to achieve this distinguished level; but what if other technologies could be utilized to aid in their quest for perfection? This project aims to apply the OpenCV library in order to track a vaulter’s movements and aid them in achieving an optimal form – the most important foundation in any sport. The athlete simply needs to provide a recorded video of their vault attempt while equipped with appropriate motion trackers. Then, through the use of motion and proximity tracking, data forecasting and transition point comparisons, a vaulter is able to receive accurate information on where he/she can improve specific aspects of his/her form. The calculations’ results are then presented in a manner that specify deviations from the optimal form through each phase and allow the vaulter to recognize their areas of improvement. Concluding recommendations are made for future work on the evolution of this tool as well as applications towards other form-specific sports.


Kaitlyn Bolton-Blevins, Rachel Hall, Tyler Shuck, Grant Napier, Sr.
"Hood College Student Information App"
Faculty Advisor: Prof. Elizabeth Chang


Josh Yurche, Sean Counterman, Ryan Bostick
"M.A.V.R.E.C. Military And Veteran Resources, Employment, Connections"
Faculty Advisor: Prof. Elizabeth Chang


SrProj2016.pngErik Phillips, Shane Edmiston
"A Novel Approach to DDoS Detection"
Faculty Advisor: Prof. A. Salem

Abstract: Denial of service attacks and distributed denial of service attacks are becoming the most common used and widespread network attack in today’s society. These attacks are affecting everything from small businesses to large companies and are occasionally the basis of threats to persuade and strong-arm large companies toward specific actions that benefit the attacker. During our research, we have studied how multiple types of denial of service attacks are created and the different methods other researchers have created to detect and prevent these kinds of attacks. Our project is aimed at creating an optimal method to sort network traffic and determine what is legitimate traffic and what is malicious in nature and consistent with denial of service attack. We will be analyzing several different algorithms for sorting using the same test data, which consists of denial of service/malicious traffic and normal traffic. In our research we found that after reading these documents that detection of malicious packets will be based on the attributes of the packets. This means that we are going to be testing the algorithms on how well they can detect normal traffic and differentiate that information from denial of service traffic. They will be separated based on the attributes of the packet information.


Kevin Barrett, Davon Hill, Devin Hill, Tyler Shaw
"Coach Central"
Faculty Advisor: Prof. E. Chang

Abstract: There exist software and mobile applications that sports coaches use in order to aid them in their everyday tasks. These softwares include certain functionality such as team management and playbook creation. However, there is no current software that includes multiple functional engines or that includes the three engines that Coach Central offers. Coach Central is created using a mySQL database, PHP, JavaScript, Ajax and is pulled together in a GUI using Bootstrap, HTML5 and CSS. Coach Central is a web and mobile accessible coaching software targeted at sports coaches. The current version of the software is available for basketball coaches only. The three main functions of Coach Central are the “My Team” engine which allows coaches to manage a current roster, the “Playbook” engine that allows coaches to create a playbook and share it with their team, and the “Recruit” engine which gives coaches the ability to manage recruiting tasks.


Antonio Punzo, Owen Rosier
"Infrastructure and Comparative Analysis of Stock Price Prediction Techniques (May 2016)"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: There are many algorithms that have shown promise when attempting to make future predictions relative to an input set of data. The motivation behind this research was to prove and implement computational methods to predicting the stock market. To answer this question, previous research had to be taken into account. The previous literature has looked into the use of prediction algorithms to analyze future trends in the stock market. With carful analysis of the literature, it was decided that the algorithms that show the most promise for predicting stock prices are Support Vector Machine, k-Nearest Neighbor, Random Forrest, and Artificial Neural Network. Linear and Logistic Regression were selected as baseline algorithms since they are simple to understand and implement in comparison to the other prediction algorithms. After implementation of each algorithm, the accuracies are compared because the methods in previous literature were done as a single, pairwise, or even theoretical comparison. This project took all evaluated methods used for computational finance and compared them in effort to see which is best for market prediction and to create a validated analysis of these predictions. The predictions were made using a static training data set and testing set across all algorithms. Once predictions were made, they were then validated by making the same predictions on another similar but different data set. The results of these predictions showed that Random Forest is one of the best algorithms for predicting the stock market, closely followed by the other algorithms with little change between prediction accuracies even when changing the data set.


Nathan Goedeke, Samuel Walters-Nevet, Joshua Williams
"ICollection and Analysis of College Student Mobility Using Wi-Fi Detection"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: In past experiments, this has been used as a safe and non-invasive means of collecting data on individuals without their knowledge. In our study, we constructed portable devices that collected information in a public area of a college campus during the spring semester of 2016. This information was then analyzed using clustering algorithms to generate for each device as well as larger trends.

Fawzya Alghamdi, David Ford, Tyler Martin
"Mining and Visualization of Political Violence Data in Africa"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The African continent has a long history of political violence. This project investigates possible correlations between the occurrence of political violence and two key quality of life variables – gross domestic product (GDP) per capita and life expectancy. Causes of political violence are discussed, and it is hypothesized that a correlation does exist between political violence and the chosen quality of life variables. The source for the political violence data is the Armed Conflict Location & Event Data Project (ACLED). Data from the World Bank was used for GDP per capita and life expectancy. Data analysis was performed using RapidMiner to determine the strength of correlation between the variables. A visualization tool was implemented using the Google Maps Application Program Interface (API) in order to visually overlay the data. Data from the analysis results indicate that there is evidence of correlation of the variables in some countries, but not for all.


Brandon Kirby, Robert Staples
"SUMO Mapmaker Add-on and Simulation Validation"
Faculty Advisor:

Abstract: Simulation of Urban Mobility (SUMO) is a software package designed to be used in order to perform traffic simulation. However, the process of using the software package itself has been shown to be complicated and complex in difficulty. The design of a mapmaker application, that is to be used in tandem with the SUMO software package when making traffic simulations, is an attempt to solve this problem. The application is designed to be friendly and easy to use while having the options and generating the correct documents to help with the most complex of traffic simulations. Using the application, two intersections were modeled in the Houston, Texas highway system, the one between I- 610 and I-10, and the one between I-610 and I-69. The intersections were then simulated with input traffic based on real world data and output traffic levels were compared to the actual real world output numbers, in order to determine the validity of the simulation. A significant flaw was discovered in SUMO’s “randomTrips” script when using specific types of maps. However, when not using that specific type of intersection, SUMO was very accurate.


Joshua J. Greer, Michael A. Mastantuono, Harryson Tun
"Cloudbase: A web enabled application for the management of aircraft flight records"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The MidAtlantic Soaring Administration (MASA) is a nonprofit aviation club that manages both engined and engineless aircraft. MASA has been using pencil and paper methods and an outdated, single user database application to report flight times associated with aircraft and pilots. Cloudbase is a mobile friendly, web based application that will replace the nondigital and existing outdated application. Pencil and paper reports and double entry of information will no longer be necessary. Cloudbase performs all the essential functions of recording and reporting aircraft and pilot flight times while providing additional functionality such as user management and reporting capabilities in a web enabled multiuser environment. Cloudbase is released under the GNU open source license enabling organizations similar to MASA across the world to modernize their record keeping.


Alexander J. Bodine, Jonathan M. Collier
"AQUA: Aquatics Web Scheduling & Registration System"
Faculty Advisor: Prof. A. Dong

Abstract: Registration and scheduling are often two time-consuming tasks. Registering clients or creating employee schedules can be particularly difficult. To address this problem, we developed a web based system that eliminates the need for paper registrations and speeds up creating worker schedules. All the data for registration and the schedule are stored on a database. The created system is being applied to the Hood College Aquatics Center, a department that currently performs the process by hand. Although the created system is customized to fit the needs of the Hood College Aquatics Center, many of the practices and methods used in the code can be manipulated to work for any sort of company with scheduling and registration requirements.


Joshua Shelley, Joshua Tokar
"Academic Assessment Data Management System"
Faculty Advisor: Prof. E. Chang

Abstract: Currently methods for recording and displaying assessment data is limited to a large spreadsheet with checkmarks to show which student outcomes, the broad goals of a course, go with each class or having reports generated by database queries with no readily available means to update and edit. For any amount of outcomes and indicators there needs to be an easier more organized way to go about these tasks. This is why we created the Academic Assessment Data Management System (AADMS). The AADMS takes all the information from available assessment data, combines it with assignment and assignment result data, and creates a user-friendly, System for professors to access their curriculum information and track their own and students' progress. Our system will be implemented in the Hood College Computer Science program. As of now Hood College uses spreadsheets spread over several web pages. There are a total of eleven student outcomes, each with two performance indicators, smaller topics that are focused on during each assignment to support the student outcome, for each outcome. Eleven student outcomes and twenty-two combined with the thirteen required classes at Hood College prove to be far too much data to be managed on a single spreadsheet. The AADMS shows all thirteen required courses and allows the professors to upload assignment information along with statistics for each assignment to create a visualization of how students are doing and in what particular student outcomes or performance indicators they may be struggling.


James Blaney Jr
"Submersible Temperature Logger"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The Brook Trout, a local trout species in MD, is facing a drastic population decline in local watersheds due to increasing water temperatures, caused by urban development and expansion. As part of an environmental rehabilitation effort, the Center for Coastal and Watershed Studies at Hood College has incubated and released several hundred Brook Trout into the surrounding Tuscarora watershed. In order to monitor the rehabilitation, the water temperature must be closely and accurately monitored. This project is composed of a hardware solution (a battery-operated, submersible temperature logger built entirely from open-source materials) and a software solution (the temperature logger firmware, a mobile application to wirelessly retrieve the data, and web site to store and present the data). An emphasis of the project was avoiding the obstacle of incorporating wireless data transfer technology (Bluetooth 4.0) into the configuration while maintaining an extended battery life for long-term deployment.

Alex Paxton
"Textbook Linguistic Measures"
Faculty Advisor: Prof. R. Ford

Abstract: This study examines various linguistic measures in college level psychology textbooks in order to determine if there has been a historical trend towards simpler sentences. The measures analyzed were the counts of words, sentences, exclamation marks, periods, question marks, word types, and a syntactic complexity estimate. These were then used to calculate the words per sentence, type-token ratio, and the mean complexity per sentence. After the data was collated, Multivariate Analysis of Variance (MANOVA) analysis was used to determine if there was a meaningful relationship, and one was found to exist. Additionally, if was determined that the words per sentence and complexity estimate both decreased over time, while the type-token ratio and the sentence count both increased over time. This indicates that sentences have gradually become shorter while the vocabulary has expanded.


Jordan McNeill, Chris Patschak, James Pool
"Raspberry Pi Web Server Load Balancing"
Faculty Advisors: Prof. G. Dimitoglou, Prof. A. Salem


Nick Aldana, Max Zarket
"Egress Simulation"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: Simulation is a powerful tool which can be used to represent and emulate real world systems. This process is becoming increasingly popular as the computational capabilities of technology become more powerful. The key point to remember when making a simulation, however, is that to validate the results, the environment needs to be as real as possible. This research is an attempt to increase the realism of agents who are evacuating from a theater by including numerous psychological and physiological factors. The results showed that some of the most important factors which influence agent egress were panic represented by choosing what exit was used, the quality of the leaders and groups, and to some degree the average space each individual had to their self. The results also showed that, in general, adding in realistic human traits and behaviors slows down the crowd’s egress times compared to homogenous, omniscient, and robotic agents.


Masha Chukhlib, Mace Stoner, Vincent Yu
"Raspberry Pi Render Farm"
Faculty Advisor: Prof. G. Dimitoglou

Abstract: The Raspberry Pi is a credit card sized, low power, low cost computer. Currently the clustering capabilities of this device have not been widely explored. Additional knowledge in this area could potentially aid in the implementation of high performance applications of the Raspberry Pi. This project is aimed at assessing the performance of a Raspberry Pi cluster. The experiment analyzes if an increase in the number of threads per node positively affects performance. The study evaluates the ability of the cluster to render a scene through Monte Carlo path tracing using variables such as a varying number of Raspberry Pi devices (nodes), threads per node, and the number of random samples per pixel used to generate the image. The data shows that, in general, adding Raspberry Pi nodes into the cluster as well as raising the number of threads per node both positively affect performance of the cluster when rendering an image through ray tracing.


Eric Diehl, Chris Fox
"Open-Source Watershed Data Logger"
Faculty Advisors: Prof. G. Dimitoglou, Prof. D. Ferrier

Abstract: Environmental research has become of growing importance. However, commercial methods of collecting environmental data are expensive and prevent many smaller organizations from owning this technology. The need is there for a low-cost alternative to these commercial systems so that small groups and organizations all around the world can have access to data about their environment as well as add to the global collection of environmental data. This project constructed an Arduino-based data logger capable of matching some of the readings that a commercial product could provide for a fraction of the cost. Data readings were successfully taken, data was pulled wirelessly from the hardware via developed software, and that data could be exported to be analyzed. This data logger is a proof of concept that the technology is now available to expand global environmental research.



Stephanie Chouinard, Travis Bellew
"mSudoku: A Mobile Puzzle Application"
Faculty Advisor: Dr. George Dimitoglou

Abstract: With the growing popularity of both Sudoku and Android smartphones, we decided to create a Sudoku application for Android 2.3.4. (Gingerbread). We strived to include helpful features that a pencil and paper puzzle couldn't provide easily or at all—like the timer, puzzle resets and hints—without overloading it or taking away the integrity of the game. The goal was a userfriendly, convenient, portable way to play Sudoku—and the app provides just that, on a small device most people already carry with them on a daily basis.


Dominick Barretta, Andrew Payne
"POV-Ray 3D Car Modeler"
Faculty Advisor: Dr. Elizabeth Chang

Abstract: There is no POV-Ray 3D car model application that is readily available in a web centric format. The lack of a user friendly platform, available anywhere, presents a problem for those that do not possess the computing power or know how to create a model car in a code driven environment. To counter this issue we constructed a mobile web platform that is used to generate a 3D model car by interfacing with the ray-tracing application POV-Ray. This platform offers features such as color, texture, background, resolution, viewing angle, and car style selections to craft a unique user designed 3D car model image and/or animation. The produced output by the application is delivered to the user via the web platform. While this platform is delineated for a distinct situation it presents a solution for the specific problem.


Jake Judson, David Lyle
"Robocall Data Mining"
Faculty Advisor: Dr. George Dimitoglou

Abstract: A "Robocall" is an automated telephone call that typically originates from a computer. The purpose is to spam as many people as possible with a message as quickly as possible using minimal effort. These types of calls are heavily regulated by the FTC but are allowed in specific circumstances, such as from political campaigns or debt collectors. Regardless, some businesses and individuals ignore the law and use Robocalls to advertise their product or scam people into providing private information, such as their credit card number, under false pretenses. Using data mining techniques, it is possible to find patterns


Andy Hohorst, Sean Kinn
Faculty Advisor: Dr. Elizabeth Chang

Abstract: JS-Ray is a web-based 3D scene builder. Its purpose is to allow a user to create a scene using a web-based real time 3D scene editor. The user can then obtain a high quality render, rendered using ray-tracing, of the 3D scene they built. JS-Ray will also allow the user to save and load their scenes on the server. The overall goal is to be able to produce high quality 3D scenes without needing to meet the system requirements for high quality 3D rendering software or installing additional software on their computer.


Tom Delaney, Jason Scaroni
"Weather Station Site & Data System"
Faculty Advisor: Dr. George Dimitoglou

Abstract: The Hood College Weather Station Site and Data System (HWS) is a follow-on project and upgrade of the Hood College Weather Station developed in 2009. This project fulfilled two goals: to re-establish HWS on new hardware and provide stability improvements, and make available more advanced tools for analyzing weather data. The goals were met by changing the existing system architecture, streamlining data processing and optimizing database storage, adding still image capture capability and creating a series of web-based dynamic graphing and analysis tools.


Andrew Blowe (BS '12), Richard Garcia (BS '12), Donald Shaner (BS '13)
"Evaluation of Hardware in the Loop Simulations of Unmanned Aerial Vehicles"
Faculty Advisor: Dr. George Dimitoglou

Abstract: The objective of this project is to compare the fidelity of a Hardware-in-the-Loop (HIL) simulation of an autonomous, fixed wing, unmanned aerial vehicle (UAV) with the results being observed in actual flight. To achieve the objective, we converted a remote controlled (R/C) aircraft to an autonomous platform and were able to perform multiple flights and collect actual data. This data was compared and analyzed against a series of simulated flights. The results showed that the HIL can be a good indicator of real life performance, yet it underscored the impact of external, environmental parameters such as the wind and idiosyncrasies of the physical platform that can only be observed during actual flights. Our findings warrant significantly more investigation and flight hours to better benchmark the performance of the physical plane and fine-tune the parameters of the simulated flights.


Allyson Cool (BS '12), Nicole Keller (BS '12), Sam Lewis (BS '12)
"Online Faculty Web Page Generator &' Templates"
Faculty Advisor: Dr. Elizabeth Chang

Abstract: With the prevalence of personal websites in today's culture it is no surprise that many people want to create and maintain an academic persona online. For people without programming experience this is a daunting task considering many of the tools available to the public today are geared towards a technological audience. For this project we created a slimmed down alternative to current content management systems (CMS). This was done by looking at our targeted audience and limiting the features offered in the CMS to the most essential tools required to make a functioning web site. These features include; creating individual web pages, linking the sites together automatically, media management, and a self contained documentation database. With these features we are able to create a lightweight CMS that provides an easy learning curve with enough customization to allow individual website creation thus matching the needs of the target audience.


Ryan Huffman (BS '12), Craig Rowe (BS '12), Sean Weber (BS '12)
"Senior Project Central"
Faculty Advisor: Dr. Elizabeth Chang

Abstract: The senior project is one of the most important requirements for a computer science degree because it requires students to compile the larger concepts and practices of their education into one project, showing what has been learned. Along with the programming and technical skills necessary to be successful, these projects require communication, organization, and resources designed for the project. Although websites such as Google Documents and Blackboard provide some of the necessary resources, no currently known website is aimed at providing specific resources for projects of this kind. Senior Project Central compiles these resources into a single website, designed to be easy to use, while allowing users to keep important project documentation in one convenient location. Developed modules such as the gantt chart creator and code execution module allow students to test and develop elements of their projects more efficiently. Senior Project Central is an essential beginning to a global solution to the organization, communication, and documentation problems associated with senior projects. With further development, it can be a permanent solution to some of the underlying issues that can affect senior projects.


Nathan Jacobson (BS '12), Naseem Zietoon (BS '12)
"Choi, the Autonomous Car"
Faculty Advisor: Dr. George Dimitoglou

Abstract: Our proposal is to adapt a R/C car to move autonomously. This project will have two main components. The first component is the conversion of the R/C Car into a robot. We are going to convert the RC car by adding an Arduino circuit board. This circuit board will allow us to control the DC motor and Servomotor. The DC motor controls the speed of the vehicle and direction forward/reverse. The servomotor controls the steering of the vehicle in degrees. The second component of the project will be to have the car be able to navigate autonomously. The car will take input from different types of sensors to handle the autonomous navigation. Based on what sensors we use, the R/C car will use the inputs from the sensors we use to make its own decisions as it navigates.


Kathleen Kaas (BS '12), Adam Stauffer (BS '12)
"Vulnerability Assessment and Reporting Tool"
Faculty Advisor: Dr. George Dimitoglou

Abstract: Vulnerability scanners are widely used as an Information Assurance tool to detect and provide remedies to computer system security vulnerabilities. The process for collecting and analyzing the results from these scanners has previously depended upon a time consuming and error prone manual process. Accordingly, there is a need for a more automated process to eliminate the possibility of human error in such critical systems. The objective of this software project is to develop a software platform and process that enables the automatic parsing and analysis of the information provided by the vulnerability scanner results. Additionally, this platform provides extensive reporting capabilities that allow the generation of short and long term trending reports from multiple vulnerability scanners. The overall platform provides a solution able to handle files from various vulnerability scanners and output the applicable data in a standardized, and easy to identify and prioritize format. The overall benefit from this software platform is an increase in productivity, efficiency and reliability in collecting, summarizing, and analyzing the output of multiple vulnerability scanners.


Ryan Kane (BS '13), Richard Orndorff (BS '12)
"SEMRS: Secure Electronic Medical Records System"
Faculty Advisor: Dr. Aijuan Dong

Abstract: SEMRS was designed specifically for the Mitchell's Plain Community Health Centre (MCHC), located in Cape Town, South Africa. MCHC's HIV/AIDS clinic caters to 2,976 patients with a current rate of 38 new enrollments for treatment per month. The existing medical record management system is paper-based and often leads to the loss or damage of patient files. This compromises the confidentiality of patient information, as well as the integrity of treatment. The objective of SEMRS is to address this problem by deploying a system that is user friendly and allows medical staff to enter patient information in real time so that records can be stored electronically and become immediately available through the internet. This project does not intend to eliminate paper files, but focuses on ensuring the safety of patient files and information to improve the overall treatment process at MCHC. This system is easy to use, and utilizes SSL encryption to ensure data safety and enforce patient confidentiality. Future work and improvements to this system could include the ability to upload existing multimodal patient?s medical records, and allow physicians and pharmacists to add/update patient prescriptions.


Jessica Baumel (BS, '11) and Corrie Myers (BS, '11)
"Interactive POV-Ray Tutorial"
Faculty Advisor: Dr. Elizabeth Chang

Abstract: POV-Ray is a free ray-tracing program that can be used to produce high-quality 3D images. Although users can consult the official POV-Ray documentation, there are no tutorials that provide hands-on learning with the program. Using a three-tiered architecture and a previously existing interactive POV-Ray tutorial as a guide, we created an interactive web-based tutorial that allows users to learn the basic functions of the program, such as lighting, textures, and creation of basic shapes. Even though it only covers the basics, the final result serves as a good foundation for future content and improvements.


Nicholas Case (BS, '11) and Daniel Thamert (BS, '11)
"Program Submission Portal"
Faculty Advisor: Dr. Elizabeth Chang

Abstract: Web-based applications that involve file transfer of files containing scripts that run on the server are susceptible to specific vulnerabilities. Current applications avoid these vulnerabilities by preventing the upload of files by file-type or by converting file-types during transfer. The goal was to design a file portal that allowed these vulnerable file-types to be uploaded while maintaining security. To achieve this, files that are uploaded are scanned for keywords defined by the administrator of the system. File-lines that contain these keywords are converted to comment lines that are not read explicitly by the server. Through this method, we developed a secure environment for file transfer that involves minimal alterations to the file and leaves the file-type intact. The security portion of the portal is designed in a modular fashion and may be implemented in larger systems; however, as input file-size and the keyword file-size increases, efficiency is sacrificed. In systems where security is more important than efficiency, this application may secure against more vulnerabilities than most.


David Cope (BS, '11) and Airrick Woolen (BS, '11)
"Solar Image Analysis Proposal"
Faculty Advisor: Dr. George Dimitoglou


Briannica Harper-Sampson (BS, '11)
"Veterans' Insurance Expenditure Database"
Faculty Advisor: Dr. Aijuan Dong


Jared King (BS, '11)
"On-line Image Rendering Tutorial"
Faculty Advisor: Dr. Elizabeth Chang


James Scott McLemore (BS, '11)
"A tasking algorithm to control a simulated autonomous robot "
Faculty Advisor: Dr. George Dimitoglou


Valentine Polii (BS, '11) and Laura Schanno (BS, '11)
"Hood College Course Assessment Site"
Faculty Advisor: Dr. Elizabeth Chang


David Blowe (BS, '11)
"Digital Reconstruction of a Roman corbita using POV-Ray"
Faculty Advisor: Dr. George Dimitoglou
Joyce Kelly (BS, '11)
"CS Technical Writing Portfolio"
Faculty Advisor: Mr. John Boon
Asma Khan (BS, '11)
"Web-based Bibliography"
Faculty Advisor: Dr. Ahmed Salem
Soofia Mujeeb (BS, '11)
"Implementing a Face Detection Program using OpenCV"
Faculty Advisor: Dr. Aijuan Dong


Jeffrey Balinsky (BS, '10)
"Elevation-based Shortest Path Algorithms"
Faculty Advisor: Dr. George Dimitoglou

Abstract: This project implemented elevation-based shortest path algorithms for street navigation on a Droid smartphone. Typical GPS applications provide the shortest paths between any two geographic points, which works well for automobile navigation, but falls short when mobility is not mechanized (e.g. riding a bicycle, walking or riding a skateboard). The project demonstrated that it is possile to generate optimal directions while considering elevation changes in the terrain.


Nicholas Burdette (BS, '10)
"A Gaming Environment for Creating Interactive Fiction"
Faculty Advisor: Dr. Elizabeth Chang

Abstract: Video game engines are used to quickly and easily create a new game by reusing code for the commonly required elements of these games. This project covers the generation of grid based maps with sprite based graphics, generation of computer controlled characters, and basic battle functions. Having these three elements allows the user to add new maps and characters to the game easier than if they had to program all of these elements from scratch. The maps and characters are generated by adding the relevant information to text files which are read by the game engine. The language used for this basic game engine is Java.


Adam D. Gilbert (BS, '10)
"Content-Based Image Retrieval"
Faculty Advisor: Dr. Xinlian Liu

Abstract: Content-based image retrieval analyzes the features of an image in order to identify similar images, which could replace the practice of finding similar images based on tags or captions, which can be time-consuming and costly to create. This study develops a system for training a system to be able to recover similar images based on the image features as well as user feedback. This system is based on systems that have been developed before that represented images based on their relevance to other images within a database. It looks to expand upon those methodologies by also accounting for the images within the database to which an image is not relevant. The study found success using this method, with one test yielding 99.81% retrieval accuracy. This study should be helpful to others looking to develop systems for content-based image retrieval.


Olabusayo A. Kilo (BS, '10)
"Digital Mammograph Classification"
Faculty Advisor: Dr. Aijuan Dong

Abstract: Because breast cancer is one of the leading public health issues today, accurate mammogram classification is an essential aspect to their survival. Unfortunately, the estimated accuracy of radiologist's mammogram classification is only 75%. Computed Aided Detection systems, using Digitized Mammograms, is hence indispensable to the field, as it increases the sensitivity of the mammogram classification. In this paper, we introduce an automated cropping mechanism and compare four image enhancement techniques, extracting intensity and texture based features from them, and running the data through a K-Neighbors classifier to evaluate their effectiveness. We specifically evaluated 4 methods of histogram equalization, Local histogram equalization, global histogram equalization, Adaptive histogram equalization and Multi-peak Generalized histogram equalization (MPGHE). The last , which was implemented by us, is not the full algorithm, but experimental results indicate it still had a favorable effect on the system. Experimental results also indicate that Adaptive Histogram Equalization fared best, followed by Global histogram equalization and MPGHE, with local histogram equalization coming in worst.


Carlin J. Rabie (BS, '11)
"Gesture to Sign Language Translator"
Faculty Advisor: Dr. Ahmed Salem

Abstract: The goal of this project was to develop an alternative way for deaf and hearing impaired people to communicate to large audiences. During presentations, deaf or hearing impaired people would be required to sit in the front most rows in order to receive the message presented by the speaker. This project would make use of the P5 glove to interpret the sign language alphabet and use a JAVA program to display the characters signed by the user. First, each letter of the alphabet was broken down in to sequences of finger bends and then a program in JAVA was written to take that character and display it in an Applet. Some of the problems I encountered were that the P5 glove was not sensitive enough to distinguish some letters with similar hand gestures. Interfacing between two unrelated things and accounting for limitations that I am not familiar with takes very creative thinking. This project could be very useful if it were implemented at any scale.


Samuel A. Stansfield (BS, '10)
"The Hood College Weather Station"
Faculty Advisor: Dr. George Dimitoglou

Abstract: The purpose of the weather station is twofold: first, to provide up to the date local information and second, to establish an easily accessible archive of weather data. The station at Hood College consists of the following components: the weather station instruments and corresponding software, the data repository and a website. An underlying data flow architecture connects these components to provide a user-friendly site that allows viewing current campus conditions, identifying extreme weather conditions (e.g. historically high and low temperatures), accessing and downloading archived data, and even providing some rudimentary dynamic charting capabilities for the visual representation of data.


Marcus Thomas (BS, '10)
"Performance Analysis Of Distributed Databases"
Faculty Advisor: Dr. Aijuan Dong

Abstract: As society and technology advance, our reliance on data grows ever more. The ability to quickly store, access, and manipulate data is becoming essential in the world today. As the focus on the reliability and availability is quickly growing, distributed databases are a way to satisfy these needs as they provide greater access and reliability of data than non-distributed databases. This project is a look at the performance analysis part of the process a company makes when evaluating the best way to house their data. The project focuses on two databases, a MySQL Cluster and a Greenplum Database. The MySQL Cluster is an open source database where as the Greenplum Database is said to be high-end. The two data databases are compared on query runtime performance of data loading, simple select queries, and aggregate queries. The results showed the load times of one gigabyte files taking up to several hours for the Greenplum Database and only minutes for the MySQL Cluster. The simple select query results showed the MySQL Cluster taking nearly a minute per transaction and the Greenplum achieving an average of around five seconds per transaction. Aggregate queries were roughly the same as the simple select query for both systems, the MySQL ran from a fifty seconds to a minute and a half for each query while the Greenplum took only seven to eight seconds. No concise conclusion could be drawn as various system restrictions impacted the results, however, the data shows that given the limitations of the environment the systems were built under, the Greenplum Database?s query performance was greater than the MySQL Cluster?s but the MySQL Cluster?s load times were far better than the Greenplum Database?s. MySQL Cluster is a good option for storing critical data that needs to be accessed quickly from anywhere if you have the resources so that the data can be stored in-memory. A Greenplum Database is a good option if you have large amounts of data (terabytes to petabytes) that need to be readily available and accessed quickly.

Earl Davis (BS, '09)
"Simulation and Analysis of Congestion and Traffic Control in Greedy Networks"
Faculty Advisors: Dimitoglou, Salem
Ryan Eaton (BS, '09)
"The Nintendo Wii and its Application Towards The Interactive Classroom"
Faculty Advisor: Salem
Kyle Huyser (BS, '09)
"superBlazer: Building Hood's First Supercomputer" (Honors Project)
Faculty Advisor: Dimitoglou
Kristopher Reese (BS, '09, BA, '09)
"Gaming Concepts in Accessible HCI for bare-hand Computer Interaction"
Faculty Advisors: Salem, Dimitoglou
Eric Walton (BS, '09)
"The Harold Weisberg Archive—Digital Collection"
Faculty Advisor: Dimitoglou


William Bonde (BS, '08)
Topic: "Feature Extraction Using Connected Component Labeling".
Faculty advisor: George Dimitoglou.
Tym Butler (BS, '08)
Topic: "Approximating a One-Time Pad Using a Genetic Algorithm".
Faculty advisors: Ahmed Salem, George Dimitoglou.
Dan Larson (BS, '08)
Topic: "A word guessing game engine using Genetic algorithms".
Faculty advisors: Ahmed Salem, George Dimitoglou.
Jason Ruvinsky (BS, '08), Paul Bartholomew (BS, '08)
Topic: "Cube World: Game Design Using A 3D Engine".
Faculty advisor: Raymond Myers.
Adam Tolley (BS, '08)
Topic: "Robotics Platform Development".
Faculty advisor: George Dimitoglou.
Eric Whitenton (BS, '07)
Topic: "A Novel Peak Detection Algorithm Used in the Study of Machining Chip Segmentation".
Faculty advisor: Ahmed Salem.
Additional info: Project results were published in CAINE 2007 - the 20th International Conference On Computer Applications In Industry And Engineering of the International Society for Computers and their Applications (ISCA).
Mayuran Thurairatnam (BS, '07)
Topic: "Lighting Using Spherical Harmonics".
Faculty advisors: Elizabeth Chang, George Dimitoglou.
Cory Petosky (BS, '07)
Topic: "Open 2Dimensional Development (o2d) Project".
Faculty advisor: George Dimitoglou.