Text File

Training Mode also requires a plain text file with the same basename as the selected image. This means that the image foo.png must have its text file named foo.txt; bar.txt will not work.

This text file should contain an ASCII representation of the image it goes with. The representation should be of the same form as the image. If the image has four lines of five words each, the text file should also have four lines of five words each. Each word in the text file should match up with the word in the same location as the image.

When being trained, the software will take the word in line 1, position 1 as being the ASCII representation for the word in line 1, position 1 of the image. Line 1, position 2 in the text file will map to line 1, position 2 in the image, etc. What ever is in the text file will be placed in the Data Store so this file should be the correct representation. If a word is spelt incorrectly then it the software will be trained with the incorrectly spelt word and when ever it encounters the word in an image it will output an incorrectly spelt version of it.