Figure - Table detection challenges

[egork520]

In this issue I am collecting some of the examples where LayoutParser predictions or Villa predictions are incorrect.

[egork520]

sha = 1dd66e46a7a92fa2936e7f8ec3d727acf5260697
Using recipes to make (box) annotations, layoutparser predictions are seem to be pretty noise. Here is an example of items in box_groups predicitons: Red = equation, Green = Figure, Yellow = Table, Pink = Text

[kyleclo]

@comorado can you include the commands you need to reproduce these predictions?

[egork520]

Hello @kyleclo here is a link to a code which reproduces the above predicted boxes