Imago Console Application

Distribution

The utility is called imago_console.exe in Windows distribution or simply imago_console in Linux and Mac OS X distributions. Binaries are offered in 32-bit and 64-bit versions.

Command-line parameters reference

The imago_console tool has an additional command line switches and possibilities in comparison to OCR Visual Tool.

The help message from the program is the following:

Usage: imago_console.exe [option]* [batches] [mode] [image_path] [-o output_file]

 MODE SWITCHES:
  image_path: full path to image to recognize (may be omitted if other switch is specified)
  -o output_file: save single recognition result to the specified file
  -characters: extracts only characters from image(s) and store in ./characters/
  -learn dir_name: process machine learning for specified collection
  -compare molfile1 molfile2: calculate similarity between molfiles
    -retcode: returns similarity 0..100 in ERRORLEVEL
  -test dir_name: calculate similarity score on specified test collection

 OPTION SWITCHES:
  -config cfg_file: use specified configuration cluster file
  -log: enables debug log output to ./log.html
  -logvfs: stores log in single encoded file ./log_vfs.txt
  -noexp: do not expand chemical abbreviations
  -pr: use probabilistic separator (experimental)
  -tl time_in_ms: timelimit per single image process (default is 10000)
  -similarity tool [-sparam additional_parameters]: override the default comparison method
  -pass: don't process images, only print their filenames
  -override config_string: override config by applying specified string

 BATCHES:
  -dir dir_name: process every image from dir dir_name
    -rec: process directory recursively
    -images: skip non-supported files from directory

 SHORTCUTS:
  -learnd dir_name: -learn -dir dir_name -images

The simplest use case is imago_console image.png which recognizes image and saves output to molecule.mol.

  • You can easily process multiple images from same directory by specifying the -dir switch.

  • The -rec switch allows recursively process and the -images switch skips all non-image data from directories. For example: imago_console -dir C:\test -rec -images

  • The -config switch allows to select configuration cluster manually, for instance imago_console -config indigo_render.txt image_rendered_by_indigo.png

  • The -log switch enables debug logging to log.html file (the required images will be stored in htmlimgs dir). The -logvfs stores all log data in single file logvfs.txt

  • The -compare switch allows to compare two molfiles and print the similarity value. If the -retcode switch specified also the result will be returned as error level variable (can be obtained under Windows by echo %ERRORLEVEL% command)

  • The -learn switch performs the machine learning process. Before you start you need to create the test collection: folder (subfolders are allowed too) where the original images and corresponding recognized structures are placed (for instance: image1.png, image1.mol, image2.jpg, image2.mol, …). The program will try to generate the configuration cluster which gives the better recognition rate on the specified test collection. If so, the cluster will be saved in current folder as config_NNNOK_MMM.txt. You can use this config further by specifying the -config config_NNNOK_MMM.txt switch.

  • The -test switch allows quickly check recognition result on a specified test collection (organized the same way as for machine learning) and returns count of correctly recognized images and average time. Can be combined with -retcode switch to return average recognition rate as process return code.

  • The -tl switch can be used for limiting the recognition time per image. If the recognition process takes more time than specified - the process will be stopped.

  • The -override switch can be used for setting config parameters. For example, -override "prefilterCV.MinGoodPixelsCount=30;" drops all connected segments with pixel count less than 30.

  • The -o switch sets the output filename for single image recognition.