Deep Learning Driven Clothing Virtual Model Generation and Animation Algorithm

Traditional clothing displays usually rely on physical models or static images, lacking interactivity and flexibility. This paper uses the StyleGAN architecture to learn human morphology and texture features from preprocessed human scan data, and also uses the CNN model to estimate human posture. Skeletal animation is constructed for the generated virtual model, the bone structure, joints and their connection relationships are defined, and the rendering technology in computer graphics is used to combine the skeletal animation with the 3D model of the virtual model to generate a smooth animation effect. This method is superior to traditional rule-based methods in terms of generation speed, realism, and motion fluency. Specifically, the reduction in generation time, the improvement in SSIM value, and the increase in motion fluency score from 7.6-9 to 9.2-10 have improved the effect of virtual model display and provided strong technical support for the digital transformation of the clothing industry.