AI Style Assist

Inspiration

A recent study done by Global Scientific Journals describes the phenomenon of how one’s sense of style affects one's psychological state. In particular, the study describes how this effect is embodied cognition in action: “Our bodies are not simply passive vessels for the mind; they actively interact with the environment, influencing how we think and feel. Clothing, in this sense, becomes an extension of the self, influencing our perception of ourselves and the world around us”. This quote describes the correlation between one’s style and their mood, proving how putting effort into one’s outfit as a form of self-care can positively impact one's life. Our team personally experiences this in our day-to-day lives. Although tempting to live the CMS stereotype and wear sweats over something nice, feeling well-dressed significantly increases our self-confidence, leading to higher levels of productivity. Although seemingly simple, choosing an outfit can feel dreadful, especially with decision fatigue. We wanted to help people avoid making more decisions and feel good about their outfit, even when they do not feel good about themselves. Thus, we came up with AI Style Assist, hoping to improve the livelihoods of people of all ages and ethnicities, with a scalable and user-friendly AI stylist. https://www.globalscientificjournal.com/researchpaper/INVESTIGATING_THE_PSYCHOLOGICAL_EFFECTS_OF_CLOTHING_CHOICES_ON_WEARER_S_MOOD_CONFIDENCE_AND_BEHAVIOUR.pdf

What it does

AI Style Assist enables you to upload images of clothing to create a digital representation of your closet. Features include filtering by clothing type, colour, and formality, a chatbot specializing in outfit generation using clothes you already have, and the ability to save favourite outfits.

How we built it

Front end: Typescript, CSS, and the framework Shadcn Database: Queried and created tables in Supabase Back end: Python, specifically FastAPI for the framework AI Portion: Pydantic AI for the agent framework Open Router for the API gateway Gemini 2.5 Flash for AI Model

Challenges we ran into

We ran into numerous challenges throughout this project. For the front-end developers, learning how to use and understand Typescript and React was challenging. But after dedicating over 35 hours, using it became a smoother journey. On the other hand, the back-end developers took a much different route, thinking more about the logistics behind our application, specifically deciding how to define style and how to measure closeness when replicating outfits.

Accomplishments that we're proud of

We are most proud of our teamwork and determination. With most of the members being first-time hackers and outside of the computer science field, we were able to build a meaningful project. In ways we lack technical skill, we make up in hard work. Deciding to step out of our comfort zone to participate in this wonderful opportunity allowed us to accelerate our learning and build things we would never be able to learn on our own. Regarding our technical process, we are most proud of the structure we created for the application. We drew out our UI on paper and created the schema for the clothes using the Buckets feature in Supabase. From there, we determined three easy steps to digitalize closets: Upload → Sort and remove background → Generate outfit. Although it does not seem like a complicated process, there were various unexpected decision-making, such as deciding which attributes to include for clothes, finding the most efficient way to generate outfits, or how we would handle layering clothes. Although under pressure, our team was communicative, eager, and optimistic, even with 1 hour of sleep. We are extremely proud of our willingness and commitment to learn!

What we learned

It was the first time Herniyan and Stella had worked on the front-end, so they had learned the entire design process in less than 48 hours. They learned how the front end is implemented and displayed, as well as how the front end and back end connect. Most importantly, they learned design workflow and how colour theory can be leveraged to enhance user-experience, through the freedom of typescript.

Lisa and Kevin, who worked on the backend, learned how to configure AI agents for the first time and learned the importance of a strong architecture/back-end structure. They learned mental execution, how to read errors, troubleshooting, debugging, and pair programming!

What's next for AI Style Assist

What’s next is creating this into a platform that is open to anyone around the globe, as well as fine-tuning our features and improving the UI to ensure smooth and safe use for its users. Here are some features we would like to improve/implement: Automatically improve the lighting of uploaded images Crop people during background removal so users can upload photos of themselves in their clothes Creating a 3D virtual model with the generated outfit

Built With

• css
• fastapi
• gemini
• openrouter
• pydanticai
• python
• shadcn
• supabase
• typescript