Prompt Refinement with Image Pivot for Text-to-Image Generation
This addresses the user experience challenge in text-to-image generation by enabling better prompt refinement without extensive parallel data, though it is an incremental improvement over existing zero-shot translation techniques.
The paper tackles the problem of automatically refining user prompts for text-to-image generation by introducing Prompt Refinement with Image Pivot (PRIP), which uses image representations as an intermediary to overcome data scarcity, resulting in substantial performance gains over baselines and effective zero-shot transfer to unseen systems.
For text-to-image generation, automatically refining user-provided natural language prompts into the keyword-enriched prompts favored by systems is essential for the user experience. Such a prompt refinement process is analogous to translating the prompt from "user languages" into "system languages". However, the scarcity of such parallel corpora makes it difficult to train a prompt refinement model. Inspired by zero-shot machine translation techniques, we introduce Prompt Refinement with Image Pivot (PRIP). PRIP innovatively uses the latent representation of a user-preferred image as an intermediary "pivot" between the user and system languages. It decomposes the refinement process into two data-rich tasks: inferring representations of user-preferred images from user languages and subsequently translating image representations into system languages. Thus, it can leverage abundant data for training. Extensive experiments show that PRIP substantially outperforms a wide range of baselines and effectively transfers to unseen systems in a zero-shot manner.