StyleXR Demo: AI-Powered Virtual Try-On in Action

Описание: StyleXR is a next-generation Virtual Try-On engine designed to revolutionize the online shopping experience. By leveraging advanced Generative AI and Latent Diffusion Models, the system allows users to visualize clothing on their own photos with hyper-realistic fidelity.
Unlike traditional overlay methods, StyleXR intelligently warps and blends garments to match the user's unique body shape and pose. It preserves critical details like fabric texture, logos, and folds while handling complex occlusions. This project directly addresses high e-commerce return rates by bridging the gap between digital browsing and the physical fitting room.
Key technical capabilities include precise human parsing using DensePose, geometric matching for accurate cloth alignment, and a customized diffusion pipeline that ensures high-resolution, photorealistic output.
Developed by Parag Bokare, Bhavesh Madhavi, Soham Patil, and Shubham Karande.

Watch StyleXR perform a seamless virtual try-on inference. This demo showcases the end-to-end pipeline:

Input: The system accepts a user photo and a target garment image.

Processing: Our Deep Learning model (based on Latent Diffusion) isolates the garment and maps it to the user's unique body pose.

Result: The output is generated in seconds, preserving the user's identity, background, and the garment's texture with high fidelity.

Tech Stack: Python, PyTorch, Diffusion Models, Gradio.

Developed by Bhavesh Madhavi, Soham Patil, Parag Bokare and Shubham Karande.