ABSTRACT
In recent years, advancements in computer vision and artificial intelligence have enabled the development of interactive applications that enhance user experience in digital environments. One such innovation is the Virtual Try-On System, which allows users to visualize clothing on their own body in real time without physically wearing the garments. This project presents the design and implementation of a web-based virtual clothing try-on platform that integrates pose estimation, image processing, and web technologies. The system aims to bridge the gap between traditional online shopping and physical store experiences by offering an immersive and personalized interface.
Conventional e-commerce platforms rely on static product images and size charts that often fail to represent how clothing fits individual users. This limitation results in uncertainty, reduced customer confidence, and increased product returns. The proposed system addresses these challenges by overlaying digital clothing assets onto a live video stream captured through a webcam. By aligning garments with body posture and movements, the system enhances visualization accuracy and decision-making during online shopping.
The application is developed using Flask for backend web services, OpenCV for real-time image processing, and MediaPipe Pose for detecting human body landmarks. Key points such as shoulders, hips, and legs are extracted to calculate body dimensions and dynamically adjust clothing size and position. This pose-based scaling mechanism ensures realistic placement of garments, including shirts, pants, and traditional outfits, even during user movement.
To improve visual realism, alpha blending and background removal techniques are applied to clothing assets to maintain transparency and smooth integration with the video feed. User authentication and session management are implemented using a SQLite database with secure password hashing to ensure controlled access. The system also supports multiple clothing selections and efficient camera control for better usability.
Overall, the Virtual Try-On System demonstrates the practical integration of AI-driven computer vision with web deployment, offering an interactive solution for digital retail. It highlights the potential for reducing dependency on physical trials while improving user satisfaction. Future enhancements may include size recommendation algorithms, 3D garment simulation, and mobile platform integration.

INTRODUCTION
In recent years, rapid advancements in computer vision and artificial intelligence have significantly transformed the way users interact with digital platforms. One such emerging application is the Virtual Try-On System, which enables users to visualize clothing and accessories on their own body in real time using a camera, without physically wearing them. This technology bridges the gap between online shopping and in-store experience by providing an interactive, personalized, and contact-free solution. The proposed system focuses on implementing a real-time virtual clothing try-on platform using computer vision techniques, pose estimation, and image processing.
Traditional online shopping systems rely heavily on static images and size charts, which often fail to provide accurate visualization of how garments will appear on individual users. This limitation leads to dissatisfaction, increased product returns, and lack of user confidence while purchasing apparel online. The Virtual Try-On System addresses these challenges by overlaying digital clothing assets directly onto the user’s live video feed, aligned with the user’s body posture and movements. By doing so, the system enhances user engagement, improves decision-making, and reduces uncertainty in online fashion selection.
The developed system is implemented as a web-based application using the Flask framework, integrated with OpenCV for image processing and MediaPipe Pose for real-time human pose detection. The application captures live video through a webcam, detects key body landmarks such as shoulders, hips, and legs, and dynamically adjusts the size and position of selected clothing items accordingly . This ensures that virtual garments such as shirts, pants, and traditional wear (chudi) are realistically placed on the user’s body.
A key feature of the system is its real-time pose-based alignment mechanism. MediaPipe Pose provides accurate landmark detection even during movement, enabling the virtual clothes to follow the user’s posture smoothly. Based on the detected landmarks, the system calculates parameters such as shoulder width, hip width, and leg length to determine appropriate scaling of clothing assets. This dynamic adaptation enhances realism and usability compared to static overlay methods.
To further improve visual quality, the system incorporates background removal and alpha blending techniques. Clothing assets are preprocessed to ensure transparent backgrounds, allowing seamless blending with the live camera feed. The overlay algorithm carefully handles transparency and boundary conditions to avoid visual artifacts. This results in a more natural appearance of garments over the user’s body, even under varying lighting conditions.
User authentication and session management are also integrated into the system to provide a personalized and secure experience. A SQLite database is used to store user registration and login credentials securely with hashed passwords. Only authenticated users are allowed to access the virtual try-on functionality, ensuring controlled usage of the application. This design makes the system suitable for deployment as a real-world web service rather than a standalone demo application.
The system supports multiple clothing categories and dynamically loads available assets from the server. Users can select different garment types, and the system instantly updates the live video feed to reflect the chosen item. Additionally, the application includes controls to start and stop camera usage, improving resource management and user control.
Overall, the proposed Virtual Try-On System demonstrates how modern computer vision techniques can be effectively combined with web technologies to build an interactive and practical solution for the fashion and retail domain. The project highlights the potential of AI-driven visualization systems in enhancing user experience, reducing dependency on physical trials, and paving the way for future innovations such as size recommendation, 3D garment modeling, and mobile deployment.

OBJECTIVES
1. To design and develop a real-time virtual try-on system using computer vision techniques.
2. To detect and track human body landmarks accurately using pose estimation algorithms.
3. To enable users to visualize different clothing items on their body through a live webcam feed.
4. To dynamically scale and align virtual garments based on the user’s body proportions and posture.
5. To implement seamless overlay of clothing assets using image processing and alpha blending techniques.
6. To develop a web-based application interface for easy user interaction and garment selection.
7. To provide secure user authentication and session management for controlled system access.
8. To support multiple clothing categories such as shirts, pants, and traditional wear.
9. To ensure smooth real-time performance without the need for specialized hardware or sensors.
10. To create a scalable and extensible system that can be enhanced with future features like mobile deployment, size recommendation, and 3D visualization.