A Framework for Portrait Stylization with Skin-Tone Awareness and Nudity Identification
This addresses practical challenges in portrait stylization for real-world applications, though it is incremental as it builds on existing Stable Diffusion methods.
The study tackled the problem of portrait stylization by developing a framework that filters harmful content and preserves skin-tone characteristics, resulting in successful real-world deployment with good performance in explicit content filtering and accurate skin-tone representation.
Portrait stylization is a challenging task involving the transformation of an input portrait image into a specific style while preserving its inherent characteristics. The recent introduction of Stable Diffusion (SD) has significantly improved the quality of outcomes in this field. However, a practical stylization framework that can effectively filter harmful input content and preserve the distinct characteristics of an input, such as skin-tone, while maintaining the quality of stylization remains lacking. These challenges have hindered the wide deployment of such a framework. To address these issues, this study proposes a portrait stylization framework that incorporates a nudity content identification module (NCIM) and a skin-tone-aware portrait stylization module (STAPSM). In experiments, NCIM showed good performance in enhancing explicit content filtering, and STAPSM accurately represented a diverse range of skin tones. Our proposed framework has been successfully deployed in practice, and it has effectively satisfied critical requirements of real-world applications.