A Local Appearance Model for Volumetric Capture of Diverse Hairstyle