I have been trying to find a direction that will work for a fairly generalizable solution/model that can detect bounding boxes for white space in a document (image) think of a pdf that needs to be filled in, where the pdf form can come in many different layouts, though there are some patterns (name form fields with letter indicators, white space (entry field for form) that is bordered by a black line, white space that doesnt have a specific border or hard-codeable way to relate the white space to fields label… that is each white space that can have text added to it is labelled in some way but not consistent, but should be intuituve enough for properlly trained vision model to detect. I have tried Landing AI vision agent, and the popular LLMs (claude/chatGPT) i suspect i may need a well annotated dataset to then finetune a YOLO or other such model, but maybe a few examples and few shot it with Gemini etc. Any good advice will be appreciated.
