CloTH-VTON+: Clothing Three-Dimensional Reconstruction for Hybrid Image-Based Virtual Try-ON

Image-based virtual try-on (VTON) systems based on deep learning have attracted research and commercial interests. Although they show their strengths in blending the person clothing image synthesizing dis-occluded regions, results for complex-posed persons are often unsatisfactory due to limitations geometry deformation texture-preserving capacity. To address these challenges, we propose CloTH-VTON+ seamlessly integrating image-based methods strength of 3D model shape deformation. Specifically, a fully automatic pipeline is developed reconstruction using reference human model: first, matched target regions simple shaped model, then reconstructed. The reconstructed can generate very natural pose transfer, retaining textures clothes. A refinement network further refines alignment, eliminating misalignment errors estimation deformed images combined utilizing conditional generative networks in-paint areas blend them all. Experiments an existing benchmark dataset demonstrate that generates higher quality comparison state-of-the-art VTON CloTH-VTON. be incorporated into extended applications such as multi-pose guided Video VTON.