Inspiration
The inspiration for this project was my girlfriend. I would sometimes find her struggling to find the right outfit that she wants to wear for the day. I decided that I could make an app that could help her decide what she should wear.
What it does
The user must upload images of the articles of clothing in their wardrobe. Our web application will take those images along with information about the current weather, the temperature, the season, and the latest fashion trends from the web to recommend possible outfits that the user can wear based on the current weather, fashion trends, color theory, and input from the user.
These outfits will be displayed on the user using image generation so that the user can try on these outfits virtually. Users will also be able to save outfits for later.
How we built it
This application uses Gradient AI's Agent platform. An agent powered by DeepSeek R1 is given descriptions of all the items in the user's wardrobe. It also has access to a knowledge base of information about fashion trends, color theory, and articles about fashion scraped from the internet.
Using that information, the agent assembles an outfit and sends images of the clothing in that outfit to Gemini's nanoBanana image generation API. NanoBanana will then take the images of the clothing and generate an image of the user wearing the full outfit.
Challenges we ran into
Assembling the knowledge base was difficult in the very beginning and we had some issues with the built-in web scraper. It was trying to scrape too many articles from the website that we pointed the scraper at.
Image generation is also a slow process as of now, so we had to deal with this by displaying loading screens and using web sockets to communicate between the browser and the server when the generated images are ready to be shown.
Accomplishments that we're proud of
We are proud to have built a working application in so short of a time frame. Our minimum viable product accomplishes the core task of deciding an outfit and showing what the outfit looks like from start to finish.
What we learned
Initially, we expected virtual try-on to be too difficult and for NanoBanana to not fully recreate the clothing we have on hand in the image generation process. However, it turns out that Nano Banana is actually very accurate in assembling an outfit from pictures of clothing. It's possible that clothing companies can leverage future image generation technology, like Nano Banana, to allow users to virtually try
What's next for DebonAIr
We believe that this app has potential. We can find even more information about fashion and integrate it into our knowledge base. We can also further streamline the process of uploading one's closet to the application. Perhaps if we implemented accounts, your wardrobe can be saved on the account and then it can be accessed on any device.
Log in or sign up for Devpost to join the conversation.