Human Parsing Based Texture Transfer from Single Image to 3D Human via Cross-View Consistency

Firstly, obtaining ground-truth 3D textures is time-consuming and labor-intensive. Secondly, textures of invisible human body parts are difficult to predict due to only one image available and lack of information from other views at inference.