maeve, a digital closet

wear things, nowhere.

Comment

by team meowmeow

maeve's inspiration

"I have nothing to wear," Maeve exasperates.

The humour of the catchphrase lies in its irony---that Maeve does, in fact, have piles upon piles of clothes gathering dust in their closet. Yet, they feel as though there is nothing that they ultimately find satisfaction with.

This highlights an individual problem of material-based insatiability, powered by increasingly consumerist culture on both social media and in real life. Trends die as fast as they are born---throwing away your Barbie-core clothes for your Brat Summer-themed ones---leading to a dangerous cycle of hyperconsumerism, materialism, and the acceleration of fast fashion companies like SHEIN, Temu, and H&M.

maeve attempts to tackle this problem on an individual level by facilitating appreciation for items we already own, promoting underconsumption and clothing re-use through artificial intelligence-powered technology.

functionality

maeve is a digital closet powered by artificial intelligence that aims to organize the user's closet so they can see the big picture of their closet and try them on virtually.

Users can upload PNG versions of their owned clothing items and classify them by brand and size. They can use our OpenCV, Streamlit-powered machine learning program to try them on virtually---no need for piles of clothes on the floor to find that matching top.

building process + frameworks

We used Streamlit and OpenCV to power the webcam-based digital fitting room. For the closet itself, we used classic React and JavaScript in a web application format, plus Tailwind and some lovely plugins for animations and PropelAuth for a login system.

key challenges

A significant challenge was implementing OpenCV and Mediapipe to enable real-time user tracking within the web app. Adjusting outfits and refining the webcam UI to align with user movements required both technical precision and artistic consideration. It took roughly 10 hours to develop efficient algorithms and loops that ensured smooth performance while maintaining the quality of the virtual closet experience. Additionally, integrating various platforms presented a complex task, demanding that each component function seamlessly across different frameworks and languages.

take-home accomplishments

  • Playing with computer vision and artificial intelligence through a comprehensive Python and Streamlit application.
  • Building snazzy animations on React.
  • Figuring out third-party authentication through PropelAuth.
  • Downing at least three cans of caffeinated soda.
  • Learning to navigate version control on GitHub.

what's next for maeve

  • Artificial intelligence-powered recommendations: given a database of clothing, artificial intelligence that recommends what goes with what.
  • Improving the accuracy of the digital fitting room: considering various body types, accessibility clauses, and cameras.
  • Building an API-based database structure for user clothing.
  • Ability to tag certain clothing items as donation-intended with attached reasons (i.e. too small.)

Built With

+ 3 more
Share this project:

Updates