screenshot-to-code-fc0aff3a·1 events·first seen Aliases: screenshot-to-code
HuggingFace introduces WebSight, a dataset designed to train vision-language models to convert web screenshots into HTML/CSS code. The dataset enables multimodal models to perform screenshot-to-code tasks, a capability relevant to UI automation and web development agents. This work represents a targeted dataset contribution for a specific multimodal grounding task.