Adding CascadeTabNet Table Structure Recognition #4
Description
Extracting table data from images is a challenge that is actively being tackled by the computer vision community. As a result, there are multiple projects and products that attempt to accomplish the task.
For its first iteration, ocirs elected to adapt the methods described in Open-Intelligence. However, the CascadeTabNet repository also includes code for applying its table detection model output towards a custom table structure recognition (TSR) algorithm.
It's currently unclear whether CascadeTabNet TSR would present a marked improvement over Open-Intelligence in terms of either accuracy or speed. This method would also only be available to users who have performed the additional setup necessary to use CasecadeTabNet.
However, it's worth at least reaching out to the developers of CascadeTabNet to compare notes on table structure recognition.