Inspiration
As teenagers, we often see many people, including ourselves, buying so many clothes from fast-fashion companies that we likely won’t use for very long anyways. While we all do want to engage in eco-friendly shopping, we’re often discouraged by the high cost associated with popular sustainable brands as well as not being able to find the styles we currently enjoy. SwapML aims to be the solution to this.
Our Team
Two years ago, 3 girls were assigned to a team as part of XSTEM, the school entrepreneurship club. Coming into the club knowing absolutely nothing about pitching, concept development, market research, or anything else about entrepreneurship, really, we can confidently say that we’ve learned more through our efforts in the club than in any classroom. This is because all our work was self-driven - we were working on exciting ideas for causes we were passionate about. Since the day we met, we’ve come so far together - we’ve traveled together, cried together, laughed together, competed and placed in numerous pitch competitions (including the World Series of Innovation and the Diamond Challenge) and, above all, become the best of friends :)) We’ve explored the entrepreneurial path separately over the years as well, through various organizations. Bhargavi has qualified for DECA Internationals and FBLA Nationals in the Entrepreneurship and Job Interview events, respectively, and also served as a Student Advisory Board member at Diamond Challenge, where she had the chance to act as a community builder between the DC team and over 5,000 global participants. She is also a fellow at YouthBridge, a diversity-oriented social impact and leadership program. Outside of this, she enjoys reading, painting, and is a horse-back rider. Aairah has competed and placed in a Marketing Series event in DECA, and is serving as an Student Advisory Board Team Chair this year at Diamond Challenge. She is also a Class Council member and on the leadership team for the school newspaper, and serves as a Content Creator on ReDefy, one of the largest student journalism platforms. You can catch her binging a Netflix show or trying out new cuisines in her free time. Diya is a talented coder and passionate researcher, currently taking several CS and science classes, including AP Comp Sci, AP Physics C, Mobile Application Development, and Virtual Reality & Game Design. She serves on the leadership team of Girls Who Code and XSTEM, and you’ll find her making (humiliating) memes of her friends or burning cookies when she’s not working on an exciting project. We have diverse, complementary skill sets that have served us well in the past - we hope we’ll be able to make use of them in the execution of SwapML!
What it does
SwapML takes an image input, compares it to the images of more sustainable options, and then returns the most similar image from our database to encourage users to purchase the sustainable version of what they already intended on purchasing.
How we built it
We first used Pandas to help us organize our dataset into folders based on the labels that were associated with each image. (e.g. all of the t-Shirts were labeled “T-Shirt” by the dataset so we put them all in a folder) We then used Google Teachable Machine to help us build a model based on Keras and PIL that classified any image by the article of clothing it was (so if we put in a shoe it will classify it as “Shoe”). We then took this information and compared the user-inputted image to all the other images in the folder of its class using a root-mean-square deviation algorithm. (the shoe gets compared to all of the other shoes in the “Shoe” folder) This algorithm returns the most similar image it found in our dataset to the inputted one.
Challenges we ran into
A challenge we ran into was certain versions of libraries not installing on Pycharm. SIFT, another component that we hope to add in the future only works with a specific version of OpenCV. However, this version was not available to install on Pycharm so we had to use another package called opencv-contrib-python instead. The SIFT component we didn’t end up weaving into our final design as we didn’t have much time but we definitely plan on integrating it into the project in the future.
Accomplishments that we're proud of
As it was our first time competing in a hackathon, we are very proud of ourselves for being able to innovate and plan an idea within the time span of 48 hours. Although the time crunch was one that we have never faced, we managed to persevere through it and put it together. We're also proud of being able to quickly put together a prototype (designing, troubleshooting) of our idea despite having limited experience in the technologies that were required.
What's next for SwapML
Next, we intend on running the SIFT algorithm on all of our images. SIFT marks important key points on images so we can compare just the key points instead of literally every image. This will reduce errors associated with the backgrounds of the clothing and do a better job at finding similar patterns. We would also like to eventually find a way to scale this to the level of the databases of entire online and brick-and-mortar (especially thrift) stores. Working on a chrome extension and a personal shopper-esque recommendation engine is in the plans as well. We have big ideas for SwapML!
Log in or sign up for Devpost to join the conversation.