Create comprehensive test suite about textrecognitiondatagenerator HOT 3 OPEN

Added some tests for the most used functions. I'll leave the issue open until I satisfied with the coverage.

Current tests are:

• String from wikipedia random
• String from test file
• String from dictionnary
• Generate image with specified format (64)
• Generate image with specified extension (.png)
• Generate image with specified skewing (15)
• Generate image with specified blur (3)
• Generate gaussian background
• Generate plain white backgroung

from textrecognitiondatagenerator.

Hi,thanks a lot for your wonderful work, there are many excellent ideas in your codes.
Something confused me, may I have a question?
1,The results are only pictures, I wonder why a file with pictures` name and label are not given. I think they are important in a OCR Deep Learning project.
2, How can I generating Samples with Text in non-Latin (English) Scripts, like Chinese or Japanese.
I have read the programs(JarveeLee/SynthText_Chinese_version, gachiemchiep/SynthText),they are not so concise and easy to understand .

from textrecognitiondatagenerator.

I am glad that you like it!

1. The label of the image is in the filename. If the image contains "TEST", the image filename will be TEST_#.jpg. If you would rather have the labels in a file I suppose that could be arranged but I see the filename as a more user-friendly (and simple to understand) solution.

2. Excellent question. I never really spent time on it because it was not my usecase. If you believe this would be an interesting feature for you I can add it.

Also, this is the issue for tests, please see #4

from textrecognitiondatagenerator.