search

DevashishPrasad / CascadeTabNet

This repository contains the code and implementation details of the CascadeTabNet paper "CascadeTabNet: An approach for end to end table detection and structure recognition from image-based documents"
MIT License
1.46k stars 422 forks source link

Result data Output Format for Border and Non-Bordered #120

Open elnazsn1988 opened 3 years ago

elnazsn1988 commented 3 years ago

Hi, Thanks very much for this, have been able to setup and replicate demo results, but am not able to understand the layout of the "Result" data. It seems that it is always an Array of size 2, with Element 1 being bbox coordinates for bordered tables, and Element 2 being Masked cooridnates for Borderless Tables.

The coordinates are in sets in 5, I was looking at border.py and it references [x1,y1,x2,y2], so what is the 5th point? Can you clarify what format you use for the bbox output, and is there a way within the code to get bbox for unbordered tables also? Much appreciated.

ghost commented 3 years ago

Hi,

The fifth point is a threshold. Based on the value u put into show_result it will plot the rectangle or not: show_result_pyplot(model, img, result, score_thr=0.85, wait_time=100)

Hope this helps!

Hi, Thanks very much for this, have been able to setup and replicate demo results, but am not able to understand the layout of the "Result" data. It seems that it is always an Array of size 2, with Element 1 being bbox coordinates for bordered tables, and Element 2 being Masked cooridnates for Borderless Tables.

The coordinates are in sets in 5, I was looking at border.py and it references [x1,y1,x2,y2], so what is the 5th point? Can you clarify what format you use for the bbox output, and is there a way within the code to get bbox for unbordered tables also? Much appreciated.

Vibha111094 commented 3 years ago

This certainly helps, however in the first element along with the result there is also somrthing that looks like below: {'counts': b'k]W38h?000000000000000000000000000000000000000000000Ub\4', 'size': [512, 512]} can we ignore it ?

Also is there a way within the code to get bbox for unbordered tables also?