The rapid rise of online shopping has contributed greatly to a paradigm shift in fashion consumer buying behavior yet, trying out clothes physically has remained a significant drawback for both offline and online fashion shopping. Customers are often uncertain about sizing, looks, and comfort of wear of garments, leading to consumer dissatisfaction and an increased return of purchased products. In offline fashion shopping, existing tryout booths in stores face constraints of limited accessibility, waiting times, and health concerns. To deal effectively with these limitations, this article proposes AI-Fit Virtual Clothing Try-On Using Deep Learning. The proposed AI-Fit system allows users to virtually view different clothes digitally without having to physically wear them using computer vision and deep learning algorithms. The system makes use of a real- time video feed obtained from a regular camera to locate the decisive points on the human body. The clothing images are then processed using deep learning algorithms to adjust them according to the user’s body movement and size as detected from the decisive points on the body. Thus, the elimination of the need for manual clothing trials achieved through the AI -Fit solution also lowers the time and effort involved, besides increasing customer confidence in buying decisions. The combination of the QR code purchasing method facilitates a smooth transition from virtual try-on to purchasing. From the perspective of retailing, the system helps in minimizing the need for the use of actual fitting rooms, in addition to reducing effort and facilitating effective handling of goods. The proposed method is thus interactive, effective, and affordable, ensuring customer satisfaction and an optimized digital buying environment. It is suggested that future improvements could include the use of gesture control.

This system overcomes two of the major limitations related to traditional stores: reliance on trial rooms and the tedious process involved in garment trials. It integrates an image capture unit using a camera with a state-of-the-art human pose estimation technique using a computer vision approach and a real-time rendering engine to provide customers with an interactive virtual environment. In this, they can visualize wearing clothes without actually putting them on. The shopping experience will be enhanced because customers can quickly compare styles, colors, and sizes without fatigue and the hassle related to multiple trials. This technology will also contribute significantly to operational benefits for retailers in terms of minimizing congestion in trial rooms, facilitating better store layouts, and being able to maintain only a limited amount of merchandise while keeping a huge digital catalog, which will improve efficiency in all respects. The system presented here does real-time overlay and fitting of garments for a single user effectively. The shortcomings in the current version are three-fold: first, it concentrates on garments related to the upper body only; second, it does not offer gesture-based interaction; and third, the support for multi-user environments is limited. In spite of these limitations, the prototype presents a scalable and pragmatic solution to bridge the gap between physical and digital retail, ensuring better purchase decisions, reduced returns, and improved customer satisfaction. Based on this, the solution of using such virtual trial solutions in modern retail in-store and online environments is recommended to enhance convenience, engagement, and operational performance. The system can be further refined with gesture-based controls to enable touchless navigation, full-body and multi-view visualization of garments to handle a wide range of garments, and accurate size recommendations by integrating body measurements with brand-specific size charts. Further integration with e- commerce can enable customers to try garments virtually at home and checkout seamlessly. Analytics may be provided to retailers with insights on usage patterns, hot items, and inventory optimization. Expanding the system for handling multiple users and smart mirror configurations may further enhance interactivity and personalization, which will ensure the evolution of the Virtual Mirror into a comprehensive, future-ready solution that transforms fashion retail into a speedier, more engaging, and technology-driven experience.

[1]H. Yang, Z. Zhang, and L. Tao, “MG-VTON: Towards multi-garment virtual try-on,” Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 1–10, 2021. [2]W. Gao, F. Hou, and Z. Deng, “CP-VTON: Clothing transfer using human parsing and pose estimation,” Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), pp. 1–8, 2018. [3]Z. Xie, S. Li, X. Dong, and F. Xu, “ClothFlow: A flow- based model for clothing try-on,” Proc. IEEE/CVF Int. Conf. Computer Vision (ICCV), pp. 1–10, 2019. [4]J. Wang, H. Liu, and Y. Yang, “Toward virtual try-on: Image-based cloth transfer using deep learning,” IEEE Access, vol. 9, pp. 55123–55132, 2021. [5]L. Ma, X. Jia, and Y. Lu, “Pose-guided human image synthesis using deep neural networks,” IEEE Trans. Image Processing, vol. 28, no. 12, pp. 6617–6630, 2019. [6]T. Chen and S. Fang, “GarmentGAN: Realistic clothing simulation using generative adversarial networks,” IEEE Trans. Multimedia, vol. 23, pp. 2471–2483, 2021. [7]M. Liu, P. Zhang, and H. Zhao, “3D virtual fitting room based on human body reconstruction,” IEEE Access, vol. 8, pp. 78934–78945, 2020. [8]M. Singh and P. S. Bedi, “Human pose tracking for virtual try-on using deep learning,” IEEE Sensors Journal, vol. 21, no. 14, pp. 15893–15903, 2021. [9]A. Rahman, Z. Uddin, and R. Islam, “Deep learning- based virtual dressing system for online shopping,” IEEE Access, vol. 10, pp. 102399–102410, 2022. [10]A. Zhalgasbayev, T. Aitken, and N. Khaimuldin, “Using EfficientNetB4 model to detect clothing in virtual try-on,” Proc. IEEE AITU Digital Generation, pp. 1–5, 2024.