search

cndplab-founder / ICDAR2019_cTDaR

The ICDAR 2019 cTDaR is to evaluate the performance of methods for table detection (TRACK A) and table recognition (TRACK B). For the first track, document images containing one or several tables are provided. For TRACK B two subtracks exist: the first subtrack (B.1) provides the table region. Thus, only the table structure recognition must be performed. The second subtrack (B.2) provides no a-priori information. This means, the table region and table structure detection has to be done.
168 stars 67 forks source link

Some problem with ctdar_measurement_tool. #4

Open AHoou opened 4 years ago

AHoou commented 4 years ago

@cndplab-founder Hi cndplab-founder,

  1. I am a postgraduate of BUPT in China. First, I am sorry to create a new issue in ICDAR2019_cTDaR to discuss a problem with the ctdar_measurement_tool，because I can't find the 'issue' button on the tool's page.

  2. when I tested and verified the function of ctdar_measurement_tool in table recognition task on the official website of ICDAR 2019 cTDaR, I found there are something wrong with it. General speaking, the question is that I found the result got by the tool was wrong when I increased the number of blank cells between two content cells. You missed to calculate the number of the blank cells in the evaluation tool, which is obviously not consistent with your method description. At the same time, ignoring changes in the number of blank cells isn’t a correct operation in the table recognition task. In order to ensure the correctness of my experiment, I can explain my verification process in detail.

  3. In order to explain the problem in detail, can you give me an email address?

IceFlameWorm commented 3 years ago

@cndplab-founder Hi cndplab-founder,

  1. I am a postgraduate of BUPT in China. First, I am sorry to create a new issue in ICDAR2019_cTDaR to discuss a problem with the ctdar_measurement_tool，because I can't find the 'issue' button on the tool's page.
  2. when I tested and verified the function of ctdar_measurement_tool in table recognition task on the official website of ICDAR 2019 cTDaR, I found there are something wrong with it. General speaking, the question is that I found the result got by the tool was wrong when I increased the number of blank cells between two content cells. You missed to calculate the number of the blank cells in the evaluation tool, which is obviously not consistent with your method description. At the same time, ignoring changes in the number of blank cells isn’t a correct operation in the table recognition task. In order to ensure the correctness of my experiment, I can explain my verification process in detail.
  3. In order to explain the problem in detail, can you give me an email address?

Hi, Is the home page of this competition still available? I tried a few times, but it always told me that the domain name of the site can not be parsed.

AHoou commented 3 years ago

I find that several months ago. And I think it has been stopped the service. @IceFlameWorm