Closet Twin: Your AI-Powered Personal Stylist Built for the Build Small Hackathon
In an era where personalization defines user experience, fashion and style remain stubbornly one-size-fits-all. That's changing with Closet Twin, an intelligent AI wardrobe assistant that transforms how people discover, organize, and wear their clothing.
The Problem
We live in a paradox: most people own more clothes than they actually wear. A typical closet contains forgotten pieces, unclear combinations, and countless moments of "I have nothing to wear"—despite a closet full of options. Style consistency is hard to maintain, wardrobes lack organization, and outfit inspiration often requires Pinterest scrolling or expensive stylist consultations.
The Solution: Closet Twin
Closet Twin bridges this gap with an intelligent, AI-driven platform that understands your wadrobe at a deeper level than any app before it. Using advanced computer vision and multimodal AI, it transforms personal style from a guessing game into a data-driven, interactive experience.
Key Features
Digital Wardrobe Management Upload and catalog your entire closet with one click Automatic attribute extraction (color, style, fit, formality, season) Smart metadata powered by MiniCPM-V-4.6 vision model Organized categorization and instant searchability
AI-Powered Outfit Generation Generate outfit recommendations based on: Weather and temperature Occasion and mood Personal style preferences Existing wardrobe inventory Interactive carousel to explore multiple outfit combinations One-click refinement: swap individual pieces without regenerating
Look Recreation from Inspiration Upload a Pinterest screenshot, Instagram post, or fashion inspiration image AI analyzes the aesthetic, colors, and energy Matches pieces from your existing wardrobe to recreate the look Identifies gaps in your closet and suggests what to buy next
Personal Analytics Track your style preferences over time Discover your most-worn colors, fits, and occasions Identify underutilized pieces and wardrobe gaps Data-driven insights to improve your closet ROI
Favorites & History Save outfit combinations for quick access Generate styled outfit cards for sharing or printing Review past recommendations and refine your preferences
Technology Stack
Built with cutting-edge open-source technology:
- Vision Model: MiniCPM-V-4.6 for intelligent image analysis
- Frontend: Gradio 6 for a responsive, intuitive UI
- Backend: FastAPI for high-performance inference
- Styling Engine: Custom carousel navigation, real-time outfit visualization
- Storage: Git LFS for efficient image management
- Deployment: HuggingFace Spaces
Why It Matters
For Users:
- Save time on outfit decisions with AI recommendations
- Maximize your clothing investment by wearing everything you own
- Discover new combinations you wouldn't have thought of Build a more intentional, cohesive wardrobe
For Fashion Tech:
- Demonstrates practical applications of multimodal AI
- Shows how LLMs can reason about subjective concepts like style
- Proves that AI can enhance rather than replace human creativity
Under the Hood
Closet Twin uses a sophisticated recommendation engine that:
- Extracts attributes from uploaded garments using vision AI
- Models your style profile from interaction history
- Generates outfits by scoring pieces against context (occasion, weather, mood, user preferences)
- Optimizes for diversity ensuring varied, interesting recommendations
- Learns from feedback to improve future suggestions
The system handles edge cases intelligently—managing incomplete data, adapting to gaps in wardrobes, and gracefully scaling from 10 pieces to 1,000+.
The Result
Closet Twin proves that AI can solve real-world problems in unexpected domains. It's not just a fun tool—it's a practical assistant that lives in your pocket, understanding your style, your schedule, and your wardrobe better than you do. Whether you're a fashion enthusiast, sustainability advocate (wear what you own!), or someone who just wants to stop thinking about what to wear, Closet Twin is reimagining personal style for the AI era.
Try it now: Closet Twin on HuggingFace Spaces Made with: Gradio • MiniCPM-V • FastAPI • Built for the Backyard AI Hackathon
Youtube demo link : https://www.youtube.com/watch?v=jkIbtqNdjx8