You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
But after training although I am getting right HTML structure but getting inaccurate coordinates result for each HTML structure even for the same image that I have used while training and looks as
Please help me to figure out why I am getting in-accurate bounding box even though the HTML structure is getting 100% accurate, Thank you so much in advance
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
I have trained the table structure Recognition model on bank statements which has a borderless table, like
For above image I have a annotation file as
{"filename": "bank_statement.png", "html": {"structure": {"tokens": ["", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", "", ""]}, "cells": [{"tokens": ["D", "a", "t", "e"], "bbox": [[10, 1], [84, 1], [84, 22], [10, 22]]}, {"tokens": ["D", "e", "s", "c", "r", "i", "p", "t", "i", "o", "n"], "bbox": [[111, 1], [723, 1], [723, 22], [111, 22]]}, {"tokens": ["A", "m", "o", "u", "n", "t"], "bbox": [[952, 1], [1014, 1], [1014, 20], [952, 20]]}, {"tokens": ["1", "1", "/", "0", "2", "/", "2", "2"], "bbox": [[11, 28], [83, 28], [83, 72], [11, 72]]}, {"tokens": ["S", "q", "u", "a", "r", "e", " "," "," ", "I", "n", "c", "D", "E", "S", ":", "2", "2", "1", "1", "0", "2", "P", "2", " ", "I", "D", ":", "L", "2", "0", "8", "7", "6", "1", "2", "4", "3", "7", "7", "7", " ", "I", "N", "D", "N", ":", "B", "u", "n", "a", " ", "E", "n", "t", "e", "r", "p", "r", "i", "s", "e", " ", "I", "n", "c", " ", "D", "B", " ", "C", "O", "I", "D", ":", "9", "4", "2", "4", "3", "0", "0", "0", "0", "2", " ", "C", "C", "D"], "bbox": [[109, 29], [725, 29], [725, 73], [109, 73]]}, {"tokens": ["2", "9", "9", ".", "7", "6"], "bbox": [[952, 29], [1019, 29], [1019, 75], [952, 75]]}, {"tokens": ["1", "1", "/", "0", "3", "/", "2", "2"], "bbox": [[10, 82], [82, 82], [82, 123], [10, 123]]}, {"tokens": ["S", "q", "u", "a", "r", "e", " ", "I", "n", "c"," "," "," ", "D", "E", "S", ":", "2", "2", "1", "1", "O", "3", "P", "2", " ", "I", "D", ":", "L", "2", "0", "8", "7", "6", "1", "5", "0", "2", "1", "1", "1", " ", "I", "N", "D", "N", ":", "B", "u", "n", "a", " ", "E", "n", "t", "e", "r", "p", "r", "i", "s", "e", " ", "I", "n", "c", " ", "D", "B", " ", "C", "C", "I", "D", ":", "9", "4", "2", "4", "3", "0", "0", "0", "0", "2", " ", "C", "C", "D"], "bbox": [[110, 82], [725, 82], [725, 130], [110, 130]]}, {"tokens": ["5", "5", "7", ".", "7", "2"], "bbox": [[959, 82], [1023, 82], [1023, 129], [959, 129]]}, {"tokens": ["1", "1", "/", "0", "4", "/", "2", "2"], "bbox": [[9, 136], [82, 136], [82, 168], [9, 168]]}, {"tokens": ["S", "q", "u", "a", "r", "e", " ", "I", "n", "c"," "," "," ", "D", "E", "S", ":", "2", "2", "1", "1", "0", "4", "P", "2", "I", "D", ":", "L", "2", "0", "8", "7", "6", "1", "8", "3", "3", "4", "5", "1", "I", "N", "D", "N", ":", "B", "u", "n", "a", " ", "E", "n", "t", "e", "r", "p", "r", "i", "s", "e", " ", "I", "n", "c", " ", "D", "B", "C", "O", "I", "D", ":", "9", "4", "2", "4", "3", "0", "0", "0", "0", "2", " ", "C", "C", "D"], "bbox": [[109, 136], [727, 136], [727, 177], [109, 177]]}, {"tokens": ["5", "2", "6", ".", "1", "6"], "bbox": [[960, 137], [1023, 137], [1023, 175], [960, 175]]}]}, "gt":"
I drew the annotated bounding box and it looks as
But after training although I am getting right HTML structure but getting inaccurate coordinates result for each HTML structure even for the same image that I have used while training and looks as
Please help me to figure out why I am getting in-accurate bounding box even though the HTML structure is getting 100% accurate, Thank you so much in advance
Beta Was this translation helpful? Give feedback.
All reactions