DreamStruct: Understanding Slides and User Interfaces via Synthetic Data Generation
This work aims to improve accessibility for people with disabilities by enabling machines to understand structured visuals, reducing the need for manual data collection and annotation.
This paper addresses the challenge of understanding structured visuals like slides and user interfaces by generating synthetic data. Their method uses code generation to create datasets with built-in labels, leading to performance improvements in tasks such as recognizing, describing, and classifying visual content types.
Enabling machines to understand structured visuals like slides and user interfaces is essential for making them accessible to people with disabilities. However, achieving such understanding computationally has required manual data collection and annotation, which is time-consuming and labor-intensive. To overcome this challenge, we present a method to generate synthetic, structured visuals with target labels using code generation. Our method allows people to create datasets with built-in labels and train models with a small number of human-annotated examples. We demonstrate performance improvements in three tasks for understanding slides and UIs: recognizing visual elements, describing visual content, and classifying visual content types.