VIRTUAL TRY-ON FOR CLOTHES USING OPENCV

Abstract: Virtual try-on systems are emerging as a transformative solution in the e-commerce domain by enabling users to visualize apparel digitally before purchase. This paper presents a lightweight and real-time virtual try-on framework developed using OpenCV and MediaPipe for accurate garment overlay. The system captures live video input through a webcam, detects human body landmarks, and overlays garments using affine transformations and alpha blending techniques. Unlike deep learning-based methods requiring extensive computational resources, the proposed system leverages classical computer vision algorithms to achieve real-time performance of 20–30 FPS on standard hardware. The system includes modules such as human detection, pose estimation, garment segmentation, overlay warping, and gesture-based interaction. Experimental evaluation shows high alignment accuracy and improved user engagement, with approximately 85% user satisfaction. The system also contributes to sustainability by reducing return rates in online shopping. Limitations include handling complex poses and dynamic fabric deformation. Future work focuses on integrating deep learning-based fitting models and 3D simulation for enhanced realism.

Keywords: Virtual Try-On, OpenCV, Computer Vision, Pose Estimation, Image Processing, Augmented Reality, E- commerce, MediaPipe