Computer Vision for Fashion: A Systematic Review of Design Generation, Simulation, and Personalized Recommendations

The convergence of fashion and technology has created new opportunities for creativity, convenience, and sustainability through the integration of computer vision and artificial intelligence. This systematic review, following PRISMA guidelines, examines 200 studies published between 2017 and 2025 to analyze computational techniques for garment design, accessories, cosmetics, and outfit coordination across three key areas: generative design approaches, virtual simulation methods, and personalized recommendation systems. We comprehensively evaluate deep learning architectures, datasets, and performance metrics employed for fashion item synthesis, virtual try-on, cloth simulation, and outfit recommendation. Key findings reveal significant advances in Generative adversarial network (GAN)-based and diffusion-based fashion generation, physics-based simulations achieving real-time performance on mobile and virtual reality (VR) devices, and context-aware recommendation systems integrating multimodal data sources. However, persistent challenges remain, including data scarcity, computational constraints, privacy concerns, and algorithmic bias. We propose actionable directions for responsible AI development in fashion and textile applications, emphasizing the need for inclusive datasets, transparent algorithms, and sustainable computational practices. This review provides researchers and industry practitioners with a comprehensive synthesis of current capabilities, limitations, and future opportunities at the intersection of computer vision and fashion design.