The code is to realize the table detection and character recognition in document. The method process is shown in following figure:

Installation

See INSTALL.md.

Whole Process

• Search the picture with table
• Table cutting
• Text detection
• Text recognition

Method 1

Command

 "sh run_whole_final.sh"

Parameters:

--test_data_path  

path of input images

--checkpoint_path=checkpoint/detect 

path of model weights

--output_chart_path 

primary path of output of  table detection

--output_chart_origin 

secondary path，which saves the original images with tables and json file recorded the bounding boxes of tables in images.

--output_chart_view   
 
secondary path，which saves the visualization of table detection of original images.
 
--output_nochart_path  

path that saves the original images without tables

--out_detect_path  

primary path, which saves the  recognition results of characters in tables

--out_detect_cut  

secondary path, which saves the images of table cutting and bounding boxes of characters.
  
--out_detect_view 

secondary path, which saves the visualization of character recognition in table.
  
--output_recog_path 

directory, which records the resuts of character recognition 

Method 2

Semi process mode

• put original images into data_new

• table detection

  "run_detect.sh"

• character recognition

  "run_reg.sh"

[detection + recognition]

  "run_detect_reg.sh"

• results detection results: data_new/jsondata recognition results: Result/reg.txt