is there any open model that can extract data from document with bounding boxes ?

[Tahirc1]

I've been working on extracting data from receipts in JSON format. However, models like PALIGemma only provide JSON without bounding boxes, while object detection models return bounding boxes but require training for specific document types. I'm looking for a model that takes a prompt with predefined classes as input and outputs JSON data with corresponding bounding boxes.