Inspiration

Fashion today is overwhelming — too many choices, not enough personalization. We realized that style discovery is often biased toward certain body types, leaving most people frustrated or insecure. We wanted to fix that. Fit Finder was born from a simple belief: everyone deserves to feel confident in what they wear. So we built an AI stylist that understands you — your body, your vibe, your confidence — and turns fashion discovery into self-expression, not self-comparison.

What it does

Fit Finder is an AI-powered personal stylist that combines computer vision, web intelligence, and design insight to personalize fashion for every body. Users upload a photo or describe their style — and FitFindr does the rest: ~Detects and categorizes body shape using AI vision ~ Scrapes Pinterest for trending outfits in that aesthetic ~Uses YOLO to identify and classify clothing pieces ~Applies Google Gemini to describe outfit aesthetics and generate fashion context ~Filters results by fit, color harmony, and body compatibility ~Lets users “like” or “dislike” outfits to refine recommendations In seconds, users get curated outfit ideas that actually fit their proportions, mood, and vibe

How we built it

Frontend: ReactJS + Tailwind CSS for a clean, editorial-style UI

Backend: Python (FastAPI) with a modular architecture for scraping, classification, and recommendation

AI & ML:

YOLOv8 for fashion item detection (shirts, pants, accessories, etc.)

Gemini Vision for outfit reasoning and style embeddings

CLIP embeddings for image similarity

Infra: Hosted on DigitalOcean Gradient AI for model serving

Data: Lightweight JSON datastore for MVP speed and simplicity

Flow: User → Scraper → YOLO → Gemini → Recommender → Ranked JSON → React UI

Challenges we ran into

Fine-tuning YOLO for fashion detection under strict time constraints.

Keeping the interface visually appealing yet responsive and minimal.

Integrating multiple AI systems (Gemini, YOLO, CLIP) into a single real-time pipeline.

Managing concurrent scraping + classification without breaking performance.

Ensuring visualizations for clothing pieces from web-scraping.

Accomplishments that we're proud of

Built a fully functional AI stylist MVP in less than 24 hours.

Combined visual understanding + generative reasoning into one cohesive pipeline.

Created a hierarchical fashion taxonomy to organize apparel data by category and layer.

Designed a clean, fashion-forward interface that feels premium yet accessible.

Presented an idea that unites technology, self-confidence, and sustainability.

What we learned

How to orchestrate multiple AI subsystems into one seamless workflow.

The power of clean data structures — our JSON pipeline saved hours of iteration.

Communicating a meaningful git hub cycle and delegating different tasks between team members.

Built our foundation for full stack development

What's next for Fit Finder

Integrate real fashion retail APIs (ASOS, Zara, H&M) for shoppable looks.

Add a 3D body-fit visualizer based on user-provided measurements.

Implement AI feedback learning for continuously improving recommendations.

Introduce sustainability filters to highlight eco-friendly and ethical brands.

Expand into a web extension or mobile app that acts as a real-time AI fashion advisor.

Our mission:

To make fashion inclusive, sustainable, and effortlessly personal — one outfit at a time.

Built With

Share this project:

Updates