Virtual Try-On Platform

An AI-powered fashion technology application that enables users to virtually try on clothing using advanced machine learning models.

Features

• 2D Virtual Try-On: Realistic garment try-on using Leffa (supports upper body, lower body, dresses)
• Body Generation: Create synthetic body models with SDXL based on user parameters
• Identity-Preserving Generation: Generate full-body images with InstantID that preserve user's facial identity
• AI Recommendations: Personalized outfit suggestions using Gemini 2.5 Flash + eBay search
• 3D Reconstruction: Generate 3D meshes from 2D images using TripoSR
• Smart Onboarding: Automatic head-only vs full-body detection
• Wardrobe Management: Store and organize garment collections
• Try-On History: Track and review previous try-on results

Tech Stack

Frontend

• Next.js 16.1.3 (React 19.2.3)
• TypeScript 5
• Tailwind CSS 4
• Three.js for 3D visualization
• Supabase for auth & storage

Backend

• FastAPI 0.128.5
• PyTorch 2.6.0+cu124
• Leffa (2D virtual try-on)
• TripoSR (3D reconstruction)
• SDXL (body generation)
• InstantID (identity-preserving generation)
• Gemini 2.5 Flash (AI recommendations)
• Supabase (storage & database)

Quick Start

Prerequisites

• Node.js 18+ (for frontend)
• Python 3.10.x (for backend)
• NVIDIA GPU with 4GB+ VRAM (recommended)
• CUDA 12.4 (for GPU support)

Frontend Setup

cd frontend
npm install
cp .env.local.example .env.local  # Configure environment variables
npm run dev  # Start dev server at http://localhost:3000

Backend Setup

cd backend
pip install -r requirements.txt
cp .env.example .env  # Configure environment variables

# Clone Leffa repository at project root
cd ..
git clone https://github.com/franciszzj/Leffa

# Start backend
cd backend
python main.py  # Start dev server at http://localhost:8000

Environment Variables

Frontend (.env.local):

NEXT_PUBLIC_API_URL=http://localhost:8000
NEXT_PUBLIC_SUPABASE_URL=your-supabase-url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your-supabase-anon-key

Backend (.env):

SUPABASE_URL=your-supabase-url
SUPABASE_KEY=your-supabase-service-key
GEMINI_API_KEY=your-gemini-api-key
USE_GPU=true
LOG_LEVEL=INFO
ALLOWED_ORIGINS=http://localhost:3000

Project Structure

/
├── frontend/          # Next.js application
├── backend/           # FastAPI ML service
├── Leffa/             # Leffa repository (clone here)
├── data/              # Runtime data storage
├── docs/              # Documentation
├── .kiro/             # Kiro configuration & specs
└── README.md          # This file

API Documentation

Once the backend is running, visit:

• API docs: http://localhost:8000/docs
• Health check: http://localhost:8000/health

GPU Requirements

• Minimum: NVIDIA GPU with 4GB VRAM (tested on RTX 3050)
• Recommended: 8GB+ VRAM for optimal performance
• CUDA: 12.4 (required for PyTorch 2.6.0+cu124)

CPU fallback available but 10-20x slower.

3D Reconstruction Setup

For 3D reconstruction features, see detailed setup guide:

• backend/3d/SETUP.md

Includes:

• Visual Studio 2022 Build Tools installation
• CUDA 12.4 Toolkit setup
• torchmcubes compilation from source
• SAM 2.1 and Depth Anything V2 setup

Documentation

• Product Overview
• Technology Stack
• Project Structure
• API Documentation
• Development Guide
• Error Handling Guide
• 3D Setup Guide
• InstantID Setup Guide

Development

Start Both Services

# Windows PowerShell
.\dev_start.ps1

Run Tests

cd backend
pytest tests/

Check Environment

cd backend
python scripts/verify_environment.py

Security

• JWT-based authentication via Supabase
• User data isolation enforced at API level
• File upload validation (max 10MB, images only)
• CORS protection with configurable origins

Performance

• GPU memory optimization for 4GB VRAM
• Model caching with singleton pattern
• Background temp file cleanup (15 min intervals)
• Automatic OOM detection and recovery
• Performance metrics tracking

License

[Add your license here]

Support

For issues and questions:

1. Check the documentation in .kiro/steering/
2. Review error handling guide: backend/app/core/ERROR_HANDLING_GUIDE.md
3. Check 3D setup guide: backend/3d/SETUP.md

---

Last Updated: February 15, 2026