Style Sensei

Daily outfit selection and style recommendations

Comment

Inspiration

Waking up and picking out an outfit every single day has always been an annoyance, and it's something I had always wished could be done for me. Creating an application that uses personal style and physical parameters to select outfits for me seemed like a creative way to help eliminate that friction in my morning routine while leveraging machine learning and computer vision.

What it does

When you wake up, you want to get your day started with as little hassle as possible. Choosing outfits in the morning can be frustrating and time-consuming, and the human mind only has so much decision-making power every day. Style Sensei takes the effort and stress out of your morning routine by selecting fashionable outfits each day based on your age, skin tone, gender, and the formality of outfit desired and season. Uploading pictures to the app will add them to your wardrobe, where the app analyzes the type of clothing as well as the color and creates fashionable outfits for your day. If you like the outfit, you can “favorite” it to save it to a collection, or you can add your own personal creations! There is also a shopping feature where you can browse items based on your preference from many different retailers to help find the best option.

How we built it

We used computer vision and machine learning via OpenCV, Google Cloud Vision, and U2Net + PyTorch to take input images of clothes and identify, segment, and inventory them in a sorted manner that would allow for the program to select outfits based on the parameters mentioned earlier. We also implemented an image scraper to search google for popular images and used that information for outfit recommendation.

Challenges we ran into

Getting the clothes to be properly identified and cropped was challenging us for a while. Many clothing images would only be partially cropped or be cropped. It was also both of our first times using Figma. Getting used to that design software was both confusing and very rewarding as we saw the app design mature.

Accomplishments that we're proud of

This is our first Hackathon so just submitting a project was our main goal and we are proud that just the two of us were able to submit one on time. We are also proud of learning how to leverage cloud services to run accurate pre-trained ML models, which without we wouldn't have had enough time to train models at all.

What we learned

Don’t try to stay awake the whole time. Sleep is very important. But besides that, we learned how to use Figma effectively for design mockups and how to use machine learning libraries.

What's next for Style Sensei

Increasing the efficiency and accuracy of our computer vision algorithm is one of our main goals moving forward along with implementing a complete and downloadable mobile app.

Built With

Share this project:

Updates