HCMUT students enhance virtual try-on tech, improving online shopping experience with AI and AR innovations.

Why not both?

Students from Faculty of Computer Science and Engineering at HCMUT is leveraging technology to enhance the online shopping experience. Their research focuses on optimizing virtual try-on (VTON) models, integrating Stable Diffusion and augmented reality (AR) to improve accuracy and realism in digital garment fitting.

Innovative approach to the current problem

Recognizing the potential of VTON technology, the team aims to refine the masking process—the critical step where a model identifies and overlays virtual clothing on a user’s body. Existing models often struggle with inaccuracies, especially in complex poses like crossed arms. By designing an enhanced masking model, the team ensures a more precise garment overlay, preserving key details such as skin tone, accessories, and natural body posture.

Their approach integrates deep learning techniques, including convolutional neural networks (CNN) for image analysis and pose estimation models to track body movements. Additionally, they employ Attention U-Net, which extracts detailed features of both garments and body structures at multiple levels, significantly improving output quality.

Looking ahead for Virtual Try-Ons

As VTON technology evolves, models are becoming less reliant on predefined masks. The team plans to explore this trend, aiming to develop an advanced try-on system that seamlessly adapts to various body shapes and movements.

At a recent exhibition at Vietnam National University – HCMC, Vietnam’s Prime Minister Phạm Minh Chính personally congratulated and encouraged the team, recognizing their research as part of the country’s key industrial development strategy. This acknowledgment has inspired the students to further their work, with a scientific paper already submitted to the Conference on Information Technology and its Applications (CITA).

With five months of dedicated research from September 2024 to January 2025, the team continues refining their model, pushing the boundaries of virtual fashion technology. Their efforts promise to enhance user experience, making online shopping more interactive and efficient.

SOURCE: Brand Management – Communications Office, HCMUT