10 Hacks from Terrapins: HackUMBC 2016

Published in

Terrapin Hackers

9 min readMar 8, 2016

Another weekend, another hackathon! Terrapin Hackers were back at it again with a great turnout and with some awesome hacks this weekend at HackUMBC 2016. From software hacks that analyze public sentiment to hardware hacks that can unlock doors with a secret knock, our students made them all! Among the 50+ hackers that attended from the University of Maryland this weekend and the many hacks that they submitted, this week we’ll showcase 10 hacks created by your very own Terrapin Hackers at HackUMBC 2016!

1. React — Aria Fallah, Robert Adkins

Top 10 Hack

You know what’s awesome about the internet? MEMES! Do you know what’s even more awesome than memes? DANK MEMES! But what sucks about dank memes is finding the “good” ones in the vast land of the world wide web. React aims to solve this problem by serving personalized memes to its users. “React watches you while you view images, and learns via sentiment analysis which types of images you like to see and feeds you more like these.” React uses machine learning to analyze a user’s facial reaction from their camera as they view a meme. It then uses the feedback from that reaction as a basis to suggest new memes. Essentially, React is using the power of machine learning to eliminate the “like” button and create a seamless user interface for browsing content on the web. The site works with many different parts but essentially uses two machine learning models. The first of which is analyzing user’s facial recognition. After a meme shows up on the page, the user’s laptop takes a picture of the user and sends it to Microsoft’s Project Oxford API, which is responsible for determining facial recognition. React then takes that information and maps it to the type of meme you just saw so that it can either feed you more or less of that type of meme. Cool huh?

Tools used: Javascript, CSS, HTML, Node.js, microsoft-project-oxford, clarifai

2. Twitter Critic — Nathaniel Self, Nicholas Kibbey, Jeffrey Feng, Brandon Grinkemeyer

NSA Sponser Winner

Sentiment analysis was a common interest and theme among many of the hacks at HackUMBC. Computers are really good at math but can they do more in tracking feelings and emotions between communications on the web? This Terrapin Hackers team aimed to solve this problem with their very own Twitter Critic. Twitter Critic rates the sentimentality of Twitter accounts and plots their tweets on a graph. Essentially, this allows users to see a sort of distribution of how positive or negative their favorite celebrity is on Twitter. The NSA, on of HackUMBC’s sponsors, also awarded Twitter Critic for their potential to deliver massive sentiment analysis on the internet.

Tools used: Java, Python, HTML5, Flask, PHP

3. The Electoral Roast — Alexander Kyei, Cole Alban, Nick Hays, Michael Wittner

BookHolders Sponsor Winner and Best Domain Name

The Electoral Roast aims to plot the world’s sentiment towards US presidential candidates on a global map. The Electoral Roast parses tweets that mention US candidates and categorizes the tweets as based on their overall feelings towards the candidates. It then plots this data across a map so that users can see specifically how districts, counties, and states feel about the candidates. The map even allows you to see specific tweets, giving you an inside knowledge on the geopolitics of the US. The team was proud to implement a machine learning algorithm and web app to create a useful program that citizens and politicians alike could use. This tool will allow people to see first hand how the internet feels about candidates and what citizens are saying about those candidates instead of just hearing about it from the polls. The team learned a lot about the Node.js/Express.js framework and plan to improve the site by focusing on the two major candidates during the general election.

Tools used: Node.js, Bootstrap, Python, ntlk, google-maps, Twitter, Firebase

4. ReactBook — Peiyong Zhang, Rex Ledesma

Facebook recently added a new feature to the site called Reactions, “a totally new redesigned like button.” Although this new feature allows users to express themselves in new and different ways, rather than just the original binary “like,” it does not let users to view global sentiment towards a trending topic on the web. ReactBook aims to solve this problem by creating graphs for trending search topics on Facebook. ReactBook reads a search query and then scrapes data from the top trending post on Facebook. It then aggregates the different reactions for that search item and depicts them on a graph. Essentially, this allows a person to see a variety of views on a certain trending topic, as opposed to just a single post from a single source. The team had difficulty scraping the data from Facebook and ran into many authentication issues. They also had difficulties in connecting their backend with their frontend, but are very proud of what they achieved in delivering real time data on public sentiments towards trending issues.

Tools used: Javascript, JQuery, Bootstrap, Canvas, Casperjs, Node.js, Phantomjs, Firebase

5. ITE Positioning System — Vincent Cozzo, Robert Rose, Weijian Cao, Daniel Corteville

The ITE Position System is a new and innovative solution to using GPS in buildings. GPS is arguably one of the greatest inventions of the 20th century, but lacks in its ability to provide accurate positioning data for small spaces. IPS solves this issue by using signal strength data from wireless routers to map users locations to rooms inside of buildings. The application uses the MAC addresses from wireless routers and their corresponding signal strengths to determine how far away the user is from each router. The app records this data every time a user enters an unmapped room so that future users will automatically know which room they are in. Although the team encountered several minor issues with working with the Android API, they were ultimately able to create a product that works. In the future, the team hopes to implement more complex machine learning algorithms to find a user’s location based on collected data.

Tools used: Android, Java

6. VibraLock — Yash Upadhyay, Daniel Engbert, Jon Baldauf, Marios Levi

Dan was annoyed about having to get up to open his door and let his friends into his room, so he came up with the idea of creating a custom lock that would allow his friends in without having to get up himself. Vibralock is a customly built hardware hack that attaches to a door and detects vibrations and knock patterns. The owner of the lock can set a secret knocking pattern as a key to open the lock. When a guest or the owner knocks on the door with the secret pattern, the lock is activated and the door opens. The team had difficulties in getting the mechanics of the Arduino to pull with enough force and rotate the door handle but was able to build a proof of concept that can essentially recognize any knocking pattern that the owner wants. The owner can also change the level of accuracy needed to match the pattern to increase the security of their system. In the future the team would like to get the mechanics of the lock working.

Tools used: Arduino, Vibration sensor

7. Pace — Daniel Briggs

Have you ever worked out to an awesome song during an intense workout only to be slowed down by a sad or slow song from your playlist? Pace aims to solve this problem by automatically serving you songs from your library that match the pace of your run or workout. Pace tracks your speed by analyzing movement in your phone’s accelerometer and then uses that data to generate a playlist of songs that matches a similar tempo. It was Daniel’s first time using Android so he had a huge learning curve to overcome. He was able to create a working app with a user interface but hopes to implement the backend logic in the near future.

Tools used: Android

8. Balloon Animal Maker — Gurpreet Singh

What’s a party if you don’t have balloons? More importantly, what’s a party if you don’t have party animal balloons, am I right? Unfortunately there are many disadvantaged youth who cannot have clown services to make them party animal balloons and as a result, these kids are forced to settle with round and uninteresting, egg shaped, blown up pieces of rubber. All kidding aside, Gurpreet, an avid and long term member of TH, set out to build an awesome machine that can build and shape balloons into party animals. Gurpreet was able to get a lot of the machinery working using wood, screws, power tools, servos, and an arduino and hopes to have this project ready to showcase for Maryland Day 2016.

Tools used: Arduino

9. TeamMemeDream — Brian Liu, Tyler Barrett

TeamMemeDream was built to find loving homes for memes without images #NoMemeLeftBehind. The idea behind this hack is to enter text into the program and have an automatically generated meme that best matches that text to a meme image. The team had trouble scraping and using Google’s TensorFlow API as none of the members have had much prior machine learning experience, but they are proud to have created the hack and be able to demo it in the short 24 hour hackathon. They learned a lot about web scraping and machine learning and hope to finish the project in the near future.

Tools used: TensorFlow API, Python

10. Text Adventure —Prathamesh Kotgire, Peter Sheu, Elijah Chanakira, Austin Piel

Text Adventure is role playing game about a warrior who travels a text based world in a mission to complete quests and defeat enemies in the forest. The user interface is completely text based and the user moves their player by typing commands such as \up \down and \map in the keyboard. The team used a random number generator to generate a random map as the user travels the grid. In the future the team hopes to add more features to the game such as adding new weapons and quests. Overall, the team is proud of getting a working demo in java and having built their first GUI and game.