English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

In this work, we focus on exploring explicit fine-grained control ofgenerative facial image editing, all while generating faithful facialappearances and consistent semantic details, which however, is quitechallenging and has not been extensively explored, especially under an one-shotscenario. We identify the key challenge as the exploration of disentangledconditional control between high-level semantics and explicit parameters (e.g.,3DMM) in the generation process, and accordingly propose a noveldiffusion-based editing framework, named DisControlFace. Specifically, weleverage a Diffusion Autoencoder (Diff-AE) as the semantic reconstructionbackbone. To enable explicit face editing, we construct an Exp-FaceNet that iscompatible with Diff-AE to generate spatial-wise explicit control conditionsbased on estimated 3DMM parameters. Different from current diffusion-basedediting methods that train the whole conditional generative model from scratch,we freeze the pre-trained weights of the Diff-AE to maintain its semanticallydeterministic conditioning capability and accordingly propose a random semanticmasking (RSM) strategy to effectively achieve an independent training ofExp-FaceNet. This setting endows the model with disentangled face controlmeanwhile reducing semantic information shift in editing. Our model can betrained using 2D in-the-wild portrait images without requiring 3D or video dataand perform robust editing on any new facial image through a simple one-shotfine-tuning. Comprehensive experiments demonstrate that DisControlFace cangenerate realistic facial images with better editing accuracy and identitypreservation over state-of-the-art methods. Project page:https://discontrolface.github.io/

DisControlFace: Adding Disentangled Control to Diffusion Autoencoder for One-shot Explicit Facial Image Editing