SustainaStyle

Inspiration

Fast fashion has an extremely bad impact on the environment. It speeds up the time taken to produce clothing items, so it consumes a high amount of resources and leads to pollution. It generates excessive waste and 85% of textiles end p in landfills. Cheap materials contribute to microplastic pollution in oceans. Fast fashion consumes high amounts of water and has more carbon emissions than international flights and shipping combined. Fast fashion fuels a cycle that's devastating our planet.

Like many other teenagers our age, our team was once oblivious to the toll of fast fashion and took part in consumption as a result of our ignorance. Now, aware of its environmental toll, we’re determined to expose its effects through our app and catalyze a shift toward sustainable fashion choices. SustainaStyle will drive us towards a more conscious yet still fashionable future.

What it does

SustainaStyle provides sustainable alternatives to fast fashion clothing items users upload into our app, helping users connect with their fashion sense while also promoting a clean environment. To further promote the importance of stopping fast fashion to create a greener and more sustainable planet, our app provides users with accessible volunteering opportunities catered towards their interests. In order to motivate users to continue promoting sustainability, our app utilizes gamification: giving users points for everytime they scan a clothing item to find alternatives for or take part in volunteering opportunities. Through these multiple features, SustainaStyle ensures environmental sustainability.

How we built it

First, we planned the UI of the app through Figma because we understood the importance of app ease of usage. After coming up with a thorough design plan, we went straight into coding. Utilizing Tensorflow, we programmed and trained a machine learning model that classifies clothing items into these ten categories: ankle boot, t-shirt, pants, pullover, dress, jacket, sandal, shirt, shoe, and a bag. Then, we utilized OpenCV in Python to detect the color of each clothing item to form a descriptor for each image the user would insert.

We utilized a MongoDB database to store essential user sign-in information, volunteering sites, and sustainable fashion stores sites. We then implemented a Flask server backend to update the database based off of user input and run the machine learning model on user inputted images. Then we wrote a webscraping program to find the top sustainable alternatives for the user’s inputted image.

Finally, through integration of front-end and back-end, we were able to create an app with seamless flow of data and advanced features.

Challenges we ran into

Our main challenges lied within training the clothing classifier model and webscraping. Utilizing artificial intelligence within this project allowed us to understand the importance of choosing accurate and lengthy data for better results. Finding large datasets that were beneficial specifically for our cause in such a short time of a hackathon was one of our main difficulties, resulting in having a model that was not as accurate as we wanted it to be. However, we ensured to do constant training of our data and modifying each layer within our convolutional neural network to make our training data and testing data results’ accuracy as high as possible, reaching a median accuracy of 91% in the end.

Webscraping also posed difficulties for us as we realized for each website, there was a different way of getting images from that site and there was very little consistently between sites. This resulted in us not being able to display images in our results as we had hoped, but however through lots of research we were still able to display accurate links for users to access sustainable alternatives.

Accomplishments that we're proud of

Our team holds pride in developing a successful machine learning model that accurately classifies over 90% of images it receives. We are also proud of our integration efforts between client and server as that has resulted in vastly advanced features within our app.

What we learned

This was one of the first times our team members got truly involved with learning how to integrate a client and server side within an application. We learned how to utilize our Flask server and API requests to transfer essential data such as images the users took and the resulting classifications the backend returned. We also learned how to train machine learning models from scratch, rather than normally just using already pre-trained models within our programs.

Even more, working on this app has allowed all of our members to truly realize the detrimental impacts of fast fashion on sustainability, and we are all more aware of how we can contribute more towards slow fashion, rather than the fast consumption of fashion, to help sustain the Earth.

What's next for SustainaStyle

SustainaStyle looks towards having a more developed clothing item classifier that can classify more than just the 10 previously mentioned categories. We also look towards developing our webscraping algorithm to be able to overcome the inconsistencies between sites and pull images from each website. We plan on incorporating more features such as having a style quiz and providing personalized tips to users on how they can maintain their style in sustainable manners.

Built With

• beautiful-soup
• flask
• javascript
• keras
• node.js
• python
• react-native
• tensorflow