Even prior to COVID-19, online shopping has become an unfortunate late-night habit for our team at GreenStyle. This has led to some questionable purchases… at some companies with even more questionable practices when it comes to sustainable and ethical practices. COVID-19 has only unfortunately accelerated these late-night sessions and with all of us staying at home and constantly scrolling on TikTok and other social media, we have seen an increase in the promotion of affordable, low-quality fast fashion brands, which has only contributed to the significant issue of textile waste. We wanted to create a tool that would educate individuals about this issue while providing affordable alternatives to shopping online.
What it does
GreenStyle is an online tool that allows users to input a link to an item of clothing they are interested in purchasing. Our tool identifies the item and brand, and checks if the brand has good ethical and sustainable practices. If they do not, GreenStyle suggests 3 alternative brands and items with similar price points, and style, along with their links so that users can find the item online and purchase it!
How I built it
GreenStyle was constructed with Python in three main sections: the web scraping and database management section, the machine learning aspect, and finally the interface. For web scraping Selenium and BeautifulSoup were used to gather the data that was then saved as csv files to query later. The machine learning aspect used a CNN with transfer learning from AlexNet, built with Pytorch. Finally, the interface was created using Flask and HTML. A more detailed outline of the steps taken to create this project can be seen in the slides and video below.
Challenges I ran into
Throughout this project we ran into a variety of challenges over the 36 hours. One particular issue we had to wrangle with was deciding how to manage the breadth and range of potential clothes out there. A dataset of clothing could be sorted into as few as 3 categories or a seemingly infinite number; we had to determine how to balance the difficulty of classifying data into smaller groups with the benefit of offering a more tailored and useful model.
Accomplishments that I'm proud of
One major accomplishment that we are proud of is the learning we all did through this project. Given the time limit, it’s always a scramble to get everything done. This is especially so when learning how to deal with new packages and challenges on the spot; for us this included learning to scrape dynamic sites with selenium as well as create user interfaces with flask. Beyond that, our team is proud of creating a product that we feel can do good in the world, and help others in similar situations like ours.
What I learned
Through this event our learning grew not only in terms of technical skills, but also in terms of the issue we were addressing. In our work to create a product that’s both user friendly as well as environmentally friendly we had to educate ourselves even more on this issue. I think we’re all coming out of this event with a new perspective the next time we consider shopping, and definitely a new tool to tackle fast fashion with.
What's next for GreenStyle
Our team at GreensStyle sees a green future ahead! We are interested in furthering our tool into a browser extension to increase accessibility and convenience for our users, as well as even an app to allow for real-time suggestions for sustainable alternatives to fast fashion brick and mortar stores once malls and shopping centers re-open. In terms of features to add onto our existing product, we would love to incorporate personalized style quizzes to generate better recommendations, partner with sustainable businesses even outside of clothing, and provide more information and statistics on sustainable & ethical practices. To us, GreenStyle is just the beginning of educating individuals on the benefits of shopping green!