Online shopping and in paticular, online clothing shopping, is becoming an increasingly larger industry. Although it is convenient and cheap, people often have difficulty discovering styles that they may find appealing. It is also difficult for people to know their shirt size, as they do not have a tangible item to examine. To combat these issues, we created Styled, a platform that can detect the user's shirt size using artificial intelligence and recommend and filter clothing based on a custom implemented algorithm.

To measure shirt size, we utilized a artificial intelligent library that can find the length of the shoulders. We use this value to determine the letter value by different intervals.

The recommmendation algorithm uses a content recommendation algorithm approach. Every item has "attributes" which are traits of the clothing.

For any large recommendation model, large amounts of data must be obtained. To obtain data on clothing, we used the Zappos Rapid Api. The data returns information regarding each item of clothing including: title, price, url, etc... We gather more data using webscraping, utilizing the library Puppeteer, visiting each of the url's and gathering description data. We collect the descriptions to find attribute keywords within the descriptions. All of the data for each clothing item is uploaded to firebase, where it can be loaded from any device.

The frontend is built with Next.js and Tailwindcss(DaisyUI). We utilized React-Webcam to access the camera. All data pulled from the Firebase Firestore database is cached to localstorage to prevent excess reads.

• The artificial intelligence model was quite difficult to get up and running because of the difficulty in transfering image files from the frontend to backend. The base64 encoding has a specific javascript prefix which makes the image invalid. This prefix must be removed before converting to a jpg.
• The data collection was difficult because of the weird formatting that we encountered when scraping.
• The firebase reads was excessively high, thus we figured out how to cache the data locally as to not go over the quota.

• Integrating vision artificial intelligence model on a web server
• Creating a novel algorithm that can recommend based on a users interests
• Deploying the entire project as microservices: Vercel(frontend), Render(AI), Firebase(Database)

• Creating a virtual "try on" feature that allows the user to see what they may look like with the shirt on
• Expanding to other kinds of clothing
• Increasing the amount of clothing in the database
• Add other stores and brands to the database

• Eric Zhang - [email protected]
• Jeff Zhou - [email protected]