Inspiration

  • We started out with an interest in generic decision-making based on neural nets. Eventually, we decided to pursue streamlining the online shopping process by using the Indico machine learning API.

What it does

  • Allows the user to input any online clothes store and filters the clothes based on your preferences.
  • Clean, intuitive and minimalist UX design gives the user an extraordinary shopping experience.
  • Uses state of the art machine learning algorithms that pinpoint the exact genre of clothing based on the user's selections.

How we built it

  • The front-end of the WebApp was based on the Bootstrap framework and used HTML5, CSS3, and JS.
  • We trained Indico’s Machine Learning model by first compiling a collection of 50 images, then trained it based on different genres of clothing (e.g. casual, preppy, graphical, streetwear).
  • We used Express, jQuery and node server to launch a server to scrape online shopping websites.
{
    casual: 0.223889081635783809,
    preppy: 0.7216173627610188248,
    graphical: 0.1606673354405158707,
    streetwear: 0.0406673354405158707
}

Challenges we ran into

  • Scraping images of clothing from large departmental stores such as H&M, Zara, Uniqlo, etc. Not only was it difficult to scrape images from dynamically rendered javascript web pages, but also because they were so frequently uploaded that they would cause our training model to malfunction
  • Connecting front-end to back-end. *After realizing that node.js runs server side and javascript runs client side we quickly had to learn to send requests using AJAX.
  • Sending specific HTTP requests through Express.js routes.
  • Time for EC2 Instance's IP address to propagate with the DNS.

Accomplishments that we're proud of

  • Successfully training a Machine Learning model to recognize multiple genres of clothing.
  • Using a completely new API that utilizes Machine Learning algorithms.

What we learned

  • Breaking down an idea into components and collaborating effectively in a minimal period of time.

What's next for Fashion Filtr

  • Hosting on fashionfiltr.com using AWS's EC2 virtual server.
  • Integrate price and direct links to make purchases.
  • Login for preference saving that will allow for better clothing decisions.
  • Upload capability for increased style recognition.
  • Apply the algorithm to streamline the decision-making process in more categories, ex: Artwork.
Share this project:
×

Updates