Learning Non-Linear Disentangled Editing For Stylegan

Abstract

EyeglassesGray Hair Age Hairline Original Slender Smiling Wavy Hair Makeup Sequential disentangled attribute manipulation. We show in this example how to achieve realistic, controllable, disentangled face editing. From the original image (center), we propose two opposite editing directions where only one attribute is manipulated at a time. To the right: 'slender', 'smiling', 'wavy hair' and 'makeup' and to the left: 'receding hairline', 'age', 'gray hair' and 'eyeglasses'. All results are obtained at resolution 1024 2 .

Stylized Face Sketch Extraction via Generative Prior with Limited Data

“…Because a distortion‐perception trade‐off exists when inverting an image to w + latent space [TAN*21], a method that avoids this trade‐off can reduce the dependency on the quality of inversion. For example, estimating the deep features directly during inversion, as proposed in the Feature‐Style Encoder [YNGH22], and fusing them with the generated features could be a possible solution. We leave this as an intriguing direction for future research.…”