Fit Builder

You work hard all day, why would you want to spend your free time scrolling through websites trying to find an outfit that fits your style. The Fit Builder will find the perfect outfit for you.

Comment

Inspiration

The inspiration for Fit Builder is the recommender systems that large companies like amazon use where they take your preferences and your purchase history and provide suggestions for other items you might like.

What it does

Fit Builder is an application to find affordable outfits based on user needs. In our prototype Fit Builder can take up to four input values: Occasion, max price, season, celebrity influence. Our application uses user input to prompt GPT-3 (Ideally this would be a model that is trained on a database just for clothes) for an outfit found online based on the users preferences. Fit Builder then displays an "outfit" which is a set of articles of clothing that are closely related to the user generated input.

How we built it

We built Fit Builder using the node.js framework. Initially we query to GPT-3 using the openai API with node.js to generate an outfit based on the user input. Once GPT-3 returns a suitable outfit, Fit Builder searches for each article of clothing using the Custom Search API from Google to find a link to each individual article of clothing. Lastly we found a CSS layout that had a simple but user friendly UI. Additionally, we built Python web scrapers using beautiful soup that were able to take a url and return the image of the article of clothing for that URL. Unfortunately due to the terms of service for various websites not allowing scraping or web crawling we were not able to implement that part of the project into the final application.

Challenges we ran into

A big challenge that we ran into was the term of service for various websites not allowing scraping or web crawling. We had intended for Fit Builder to display images of the articles of clothing that are displayed using the URL that are generated from the Custom Search Google API. Unfortunately, for nearly every different website our web scraper would need to generate a new header and new cookies. Given the short timeframe this was not able to be implemented into the prototype of Fit Builder but would ideally be implemented in future versions.

Accomplishments that we're proud of

We are proud that we were able to connect our web application with the openai API for GPT-3 and with the Google Custom Search API to search the web. Similarly we were able to incorporate a simple but good looking css package to give our application some character.

What we learned

Through this project we were able to learn about the react javascript framework. We learned about various components and how to generate dynamic websites on the internet. We also learned about web scraping. We tested the beautiful soup web scraper to gather images from various websites.

What's next for Fit Builder

Ideally Fit Builder would have its own database with clothing with various values for various parameters rather than relying on GPT-3 to try and generate an out. Additionally, in the future Fit Builder would be able to display images of articles of clothing that the application suggests. This would give the user the ability to really see if they would like the outfit combination that was generated or not. In the very far distance we would want Fit Builder to be able to display the outfit on a generated dummy that has similar dimensions to the user, this would give the user even better insight into how the outfit might fit.

Built With

Share this project:

Updates