Bounding boxes for multiple languages and types of text

[wendlerc]

Hello,

I am new to this project. I am working on the creation of a large-scale open-source image dataset with OCR annotations and am wondering whether PaddlePaddle can be used to generate bounding boxes for multiple types of text using only one model. Ideally, given an image containing handwritten and printed text, I would like to get bounding boxes for both and subsequently recognize the text within those boxes using several differnt models. Ideally, I would like to detect all available languages, in printed documents, natural scenes, handwritten, handwritten math, latex equations and so on.

Best,
Chris