FaceShapeGene: A disentangled shape representation for flexible face image editing

How do I look if have the same nose shape as my favorite star? Existing methods for face image manipulation generally focus on modifying predefined facial attributes, editing expressions and changing styles, where users cannot control shapes of specific semantic parts freely in generated image. The part are described by their geometries need to be controlled continuously generating geometric parameters. Therefore, existing that work with discretely labelled attributes not applicable this task. In paper, we propose a novel approach learn disentangled representation image, namely FaceShapeGene, which encodes information into separate chunks latent space. It allows recombine part-wise from other individuals transfer specified shape, just like gene editing. Experimental results several tasks demonstrate proposed FaceShapeGene correctly disentangles features different parts. Comparisons show superiority method tasks.