Controllable Textual Inversion for Personalized Text-to-Image Generation
This work addresses the problem of making personalized text-to-image generation more robust and data-efficient for users, representing an incremental improvement over existing TI techniques.
The paper tackles the challenges in Textual Inversion (TI) for personalized text-to-image generation, such as high data requirements and lack of robustness, by proposing Controllable Textual Inversion (COTI), which achieves a 26.05 decrease in FID score and a 23.00% boost in R-precision compared to prior methods.
The recent large-scale generative modeling has attained unprecedented performance especially in producing high-fidelity images driven by text prompts. Text inversion (TI), alongside the text-to-image model backbones, is proposed as an effective technique in personalizing the generation when the prompts contain user-defined, unseen or long-tail concept tokens. Despite that, we find and show that the deployment of TI remains full of "dark-magics" -- to name a few, the harsh requirement of additional datasets, arduous human efforts in the loop and lack of robustness. In this work, we propose a much-enhanced version of TI, dubbed Controllable Textual Inversion (COTI), in resolving all the aforementioned problems and in turn delivering a robust, data-efficient and easy-to-use framework. The core to COTI is a theoretically-guided loss objective instantiated with a comprehensive and novel weighted scoring mechanism, encapsulated by an active-learning paradigm. The extensive results show that COTI significantly outperforms the prior TI-related approaches with a 26.05 decrease in the FID score and a 23.00% boost in the R-precision.