Just like many people around the world I struggle with the problem of not knowing what to wear for the day. This app aims to solve that issue. It uses AI and computer vision to build an outfit for you that's stylish and weather appropriate.

What it does

It's a mobile Android application that allows for the user to upload pictures of the clothes they have and will generate outfits from them. It uses a Pytorch model along with a K-means clustering algorithm to determine the type of clothing and color. It then learns outfits from different trends and rebuilds those outfits with your clothes.

How I built it

It was built by using python to develop and train the model. The model was built using Pytorch and a library that sits on top of that called Torchvision which was used to develop the object detection model. I also built up datatables that would be used to help determine stylish clothes by having the model infer pictures of people wearing stylish clothes and filling out the tables with the color of the clothes and the type of clothing being matched. The program would then look and see what colors and clothes matched were most common and recommend those the most.

Challenges I ran into

This was my first time doing machine learning and computer vision so just learning the basics of that and how to build a solid model. It was also my first time building and android app so just learning how different layouts and screens mingle with each other was a difficult task.

Accomplishments that I'm proud of

Although I couldn't accomplish everything I wanted done, I was able to learn A TON from this project. I learned the basics of computer vision and app development which are two fields I want to get more into. I'm really proud of the fact I was able to even build a model to begin with! (haha) Also figuring out that you can use k-means clustering to determine common color groupings was something super cool!

What I learned

Learning about Pytorch and computer vision in general. Using the torchvision online tutorial I thought was especially helpful in learning. I also learned about K-means clustering and how it relates to colors. I was also stoked to learn about Android app development and more into the backend of how they work.

What's next for DressUp

DressUp still has A LOT to be done. It still needs to have the model I trained imported over and I still need to implement some more of the backend of the app. It also needs to look more visually appealing to the user as well as implementing taking into consideration the weather. I think finding out to somehow implement Facebook Marketplace and possibly have recommendations for clothes being sold on there in the app would be cool as well!

Built With

Share this project: