VirtualCloset

Our logo

Inspiration

Many people struggle to decide what to wear or simply don’t have the time to design stylish outfits. We were inspired by this everyday challenge and wanted to make fashion smarter and more effortless. Our goal was to combine creativity with technology — using AI to help people express their style with confidence and ease.

What it does

VirtualCloset lets users input the clothes they own and connects to the Gemini API to suggest what items pair best together. It can also save individual clothing pieces and the outfits it generates, allowing users to build and manage their personal digital closet.

How we built it

We built VirtualCloset using FastAPI for the backend, JavaFX for the frontend, and Google’s Gemini API for AI-powered outfit generation. Our backend handles user input, clothing data storage, and communication with Gemini through custom API routes. We used SQLite as a lightweight database and Pandas for data manipulation and management. The frontend was developed in Java, providing a clean and interactive user interface that connects directly to our FastAPI endpoints.

Challenges we ran into

One of our main challenges was connecting our local backend code with the Gemini API and ensuring smooth communication between different parts of the project. We also had to manage environment variables securely while testing locally, handle JSON data returned from the API, and make sure our frontend and backend stayed properly linked during development.

Accomplishments that we're proud of

We’re proud that our team successfully developed a working prototype of VirtualCloset, where users can describe their clothes and instantly receive AI-generated outfit suggestions. Connecting our backend with the Gemini API and creating a virtual closet capable of saving, editing, and pairing clothing items were major technical milestones. We’re also proud of how well we collaborated across backend and frontend — combining different frameworks into one smooth, functional system in a short amount of time.

What we learned

Through building VirtualCloset, we learned how to integrate an AI model like Gemini into a full-stack environment and manage structured communication between multiple services. We improved our understanding of FastAPI, data handling with Pandas, and real-time API interaction. Most importantly, we learned how teamwork, problem-solving, and creativity come together to turn an ambitious idea into a real product.

What's next for VirtualCloset

Next, we plan to expand VirtualCloset with image recognition features so users can upload photos of their clothes instead of typing descriptions. We also want to make the system more personalized by learning each user’s style preferences and suggesting new combinations automatically. In the future, we hope to turn VirtualCloset into a mobile app that helps people plan outfits, explore trends, and rediscover the clothes they already own.

Built With

Submitted to

HACK.COMS '25

Created by

I worked on the frontend, including developing the Closet page and connecting the backend to the frontend using HTTP requests and JavaFX.

It was a challenge programming on JavaFX and connecting it to the Python/Gemini/FastAPI backend since it was my first time using the library. However, throughout the process, I learned how to utilize and configure JavaFX such that it runs functions without any problem.

Aparnaa Senthilnathan
I'm Aparnaa Senthilnathan (she/her), a 5th year Computer Science BS/MS student. I'm looking forward to being more active in hackathons.
I worked on the frontend using JavaFX.
It was definitely a challenge as it was my first time doing/using either, but I now have a understanding of how frontend and backend systems interact and how JavaFX handles user input and HTTP requests.

Emma Zheng
I worked on backend, focusing on data manipulation and clothing entity management features, using Pandas and SQLite.
I also helped integrate these backend endpoints with our FastAPI app, ensuring smooth communication with the Gemini AI service for outfit generation.

Wendy Wang
CS grad student in RIT
I led the backend architecture for this project, leveraging FastAPI to build the core intelligence engine for seamless closet management (CRUD) and AI-driven fashion advice. One of my achievement was engineering the Gemini 2.5 Flash service through meticulous prompt tailoring, which was the turning point for the project, enabling the chatbot to generate highly reliable, structured, and realistic outfit ideas instead of generic responses. A challenge I overcame was implementing the API logic to accurately parse unstructured user text into structured attributes, a critical step before generating suggestions and returning this data via robust API router requests. This work provided an excellent opportunity to master new tech stacks, including Pydantic for robust data validation, Uvicorn for high-performance asynchronous serving, and gaining familiarity with SQLAlchemy concepts for future database integration, ultimately ensuring a convenient and powerful experience for users.

Irin Boksh

Updates

Aparnaa Senthilnathan started this project — Nov 02, 2025 08:56 AM EST

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.