AI-Based Virtual Try-On System with Automatic Cloth Extraction and Real-Time Body Tracking

Замовник: AI | Опубліковано: 26.02.2026

We are looking to develop an advanced Virtual Try-On system to be integrated into our existing Web application. The system should allow users to try clothes (listed in our app) in real-time using live camera video. The system must support: Automatic cloth extraction from product images (2D or 3D) Real-time body tracking using camera Cloth overlay aligned to body pivots (shoulder, waist, etc.) Dynamic movement of clothes according to user’s body motion Scalability to support multiple products in the catalog Scope of Work 1. Automatic Cloth Cropping (AI Segmentation) From product images (model images or flat-lay images), the system should: Automatically detect and extract only the garment Remove background and human model if present Output should be a clean transparent PNG or structured 3D mesh The system must work for multiple clothing categories (shirts, t-shirts, dresses, etc.) Possible technologies: AI segmentation models Computer vision-based object detection 2. 2D to 3D Conversion (If Required) If product images are 2D: Convert garment into 3D mesh suitable for try-on Define garment pivots (shoulder, neck, waist alignment points) Maintain realistic fabric alignment If products are already in 3D: Optimize meshes for real-time rendering 3. Real-Time Body Tracking Using live camera feed: Detect human body landmarks Track shoulder, waist, chest, hip, arm positions Maintain stable tracking during movement Possible reference technologies: Pose estimation models Depth-based body tracking 4. Cloth-to-Body Pivot Mapping Match garment pivot points with body landmarks Maintain proper scaling based on user body size Cloth must move dynamically with: Arm lifting Body rotation Leaning Walking 5. Live Try-On Rendering Real-time overlay of garment on live video Maintain frame rate performance (minimum 20–30 FPS) Support Android and iOS devices Optimize for mobile performance 6. Multi-Product Scalability The try-on system should work for all products listed in the app No manual pivot creation per product (preferred) If manual mapping is required, provide scalable backend tool to manage it Deliverables AI cloth extraction module Real-time body tracking module Cloth-to-body alignment engine Mobile SDK / API for app integration Admin panel (if required) for product upload & pivot adjustment Documentation & deployment support Preferred Technical Stack (Open to suggestions, but experience in the following is preferred:) AR frameworks 3D engines (Unity / Unreal) Computer Vision & AI models Web integration Real-time rendering optimization