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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Despite large-scale diffusion models being highly capable of generatingdiverse open-world content, they still struggle to match the photorealism andfidelity of concept-specific generators. In this work, we present the task ofcustomizing large-scale diffusion priors for specific concepts asconcept-centric personalization. Our goal is to generate high-qualityconcept-centric images while maintaining the versatile controllability inherentto open-world models, enabling applications in diverse tasks such asconcept-centric stylization and image translation. To tackle these challenges,we identify catastrophic forgetting of guidance prediction from diffusionpriors as the fundamental issue. Consequently, we develop a guidance-decoupledpersonalization framework specifically designed to address this task. Wepropose Generalized Classifier-free Guidance (GCFG) as the foundational theoryfor our framework. This approach extends Classifier-free Guidance (CFG) toaccommodate an arbitrary number of guidances, sourced from a variety ofconditions and models. Employing GCFG enables us to separate conditionalguidance into two distinct components: concept guidance for fidelity andcontrol guidance for controllability. This division makes it feasible to traina specialized model for concept guidance, while ensuring both control andunconditional guidance remain intact. We then present a null-textConcept-centric Diffusion Model as a concept-specific generator to learnconcept guidance without the need for text annotations. Code will be availableat https://github.com/PRIV-Creation/Concept-centric-Personalization.

Concept-centric Personalization with Large-scale Diffusion Priors