I send it in the Demo.ipynb code first, the ID of the text tags and the PARSED_CONTENT_LIST is inconsistent.

[DAAworld]

I found that the digital labels on the picture and the index of the PARED_CONTENT_LIST can be corresponding. It's just that the two are different. For example, the text of the digital label 12 is DOG, and the index in the PARSED_CONATENT_LIST is the confnt of 12. But Bbox's coordinates are consistent.