Abstract:
With the shift in the e-commerce paradigm, particularly in the fashion industry, one of the greatest barriers still remains that customers cannot experience clothing in real life prior to the purchase. It has been a barrier that often results in uncertainty regarding fit, fashion, and overall appearance, which translates to heavy returns, customer dissatisfaction, and operational inefficiency for retailers. Virtual Try addresses this challenge with the introduction of a cross-platform mobile application where users can virtually try clothing in real-time using live camera input or uploaded images.
The platform utilizes a mixed method involving the integration of Augmented Reality (AR), computer vision, and machine learning mechanisms. Major technologies involve MediaPipe Pose for pose landmark detection, OpenCV for image processing and overlay rendering, and NumPy for processing of computational transformations. Blob Detection provides isolation of regions of interest of importance, and the Kalman Filter provides temporal smoothing of pose estimation, improving visual stability and realism under interactive user input.
Early results are improved user satisfaction and reduced reluctance to purchasing fashion items online, with the potential to significantly lower product return rates. Limitations are small pose misalignment under extreme lighting or occlusion, and different AR performance across hardware tiers, pointing to areas for future optimization.
Evaluation findings show that the system has a pose detection reliability of 100% and an average frame rate of 14.26 FPS, which is adequate for smooth real-time interaction. The accuracy of garment overlay was also very high, with misalignment kept below 5%, and adaptive rotation and scaling were efficient on a variety of body poses. The system also handled edge cases gracefully through filtering rotations above ±45 degrees. These results confirm the responsiveness and feasibility of the method, with the implication of scope for performance improvement on less powerful hardware.
Real-world solutions are tremendous, ranging from scalable, immersive digital shopping experience solution to be provided to fashion retailers. The innovation of Virtual Try is to combine real-time vision technology and convenient-to-use mobile interface, providing measurable value to end-users and online fashion platforms.
