Fine-tuning for table bounding box extraction
#18
by
TheOfficialAJ
- opened
I am currently working on parsing tables as well as extracting table bounding boxes from documents. However, I noticed that the current bbox variant of the model seems to only detect images not tables.
Has anyone experimented with fine-tuning this model on a dataset of documents with labeled table bounding boxes? I am interested in knowing how it performs on that task and what are the best ways to go about it.
Hi,
The same approach we applied to image bounding boxes can be extended to table bounding boxes given the proper data. We did not feel the need to include tables beacuse we can cleanly extract them as HTML tables.