All terms
Multimodal
Image to Text
Producing text from an image, such as a caption or extracted content.
Definition
Image to text is the task of producing text from an image, covering captioning, answering questions about a picture, and reading text out of it. An image reader picks out what is in the picture and a language model turns that into words. It is a core capability of models that combine vision and language, and a building block for document and accessibility applications.