Problem
Research questions and friction points this paper is trying to address.
E-commerce Poster
Visual Text Generation
Poster Editing
Unified Framework
PosterText: Towards Unified Visual Text Generation and Editing for E-commerce Poster
🤖 AI Summary
This study addresses the challenge of unifying e-commerce poster generation and precise editing by proposing PosterText, a framework that treats text blocks as atomic units to establish a unified generative-editing paradigm. The model is trained via a four-stage curriculum learning strategy comprising text rendering pre-training, instruction tuning, preference alignment, and spatially-guided self-distillation. Accompanied by a newly constructed large-scale dataset and benchmark, extensive experiments demonstrate that PosterText achieves state-of-the-art performance in poster generation, text block manipulation, and style control. These results effectively validate the proposed unified paradigm’s superiority in simultaneously ensuring high generation quality and editing flexibility, offering a robust solution for complex visual design tasks.
📝 Abstract
Automated e-commerce poster design requires both high-quality poster generation and flexible editing of existing designs. However, most existing methods either target end-to-end poster generation or follow multi-stage design pipelines, with limited capability for flexible and precise editing of existing posters. To enable unified generation and editing of e-commerce posters, we introduce Text Patch Generation and Editing, a unified task formulation that treats text patches as atomic units and covers four operations: poster generation, patch addition, patch deletion, and patch modification, with optional reference-guided style control. Based on this, we propose PosterText, a unified model trained with a four-stage curriculum, including text rendering pretraining, instruction-following training, reinforcement learning for preference alignment, and spatial guidance self-distillation for execution refinement. We further construct a large-scale dataset with patch-level annotations and a comprehensive benchmark for evaluation. Extensive experiments demonstrate that PosterText achieves competitive performance against existing generation and editing approaches, validating the effectiveness of the proposed framework.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Unified Visual Text Generation and Editing
Text Patch Generation and Editing
PosterText
Four-stage Curriculum Learning
Spatial Guidance Self-distillation
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
