TextRecognitionDataGenerator

A synthetic data generator for text recognition

What is it for?

Generating text image samples to train an OCR software

What do I need to make it work?

I use Archlinux so I cannot tell if it works on Windows yet.

Python 3.X
OpenCV 3.2 (It probably works with 2.4)
Pillow
Numpy
Requests
BeautifulSoup

You can simply use pip install -r requirements.txt too.

How does it work?

python run.py -w 5 -f 64

You get 1000 randomly generated images with random text on them like:

What if you want random skewing? Add -k and -rk (python run.py -w 5 -f 64 -k 5 -rk)

But scanned document usually aren't that clear are they? Add -bl and -rbl to get gaussian blur on the generated image with user-defined radius (here 0, 1, 2, 4):

Maybe you want another background? Add -b to define one of the three available backgrounds: gaussian noise (0), plain white (1) or quasicrystal (2).

Or maybe you are working on an OCR for handwritten text? Add -hw! (Experimental)

It uses a Tensorflow model trained using this excellent project by Grzego.

The project does not require TensorFlow to run if you aren't using this feature

The text is chosen at random in a dictionary file (that can be found in the dicts folder) and drawn on a white background made with Gaussian noise. The resulting image is saved as [text]_[index].jpg

New

• You can "fake" handwriting using -hw
• You can add gaussian blur to the resulting image
• Sentences from Wikipedia can be used instead of random words with python run.py -wk 1 (requires an Internet connection)
• Sentences can be picked from a file passed as a parameter with python run.py -i ./texts/random_1.txt

There are a lot of parameters that you can tune to get the results you want, therefore I recommand checking out python run.py -h for more informations.

Can I add my own font?

Yes, the script picks a font at random from the fonts directory. Simply add / remove fonts until you get the desired output.

It only supports .ttf for now.

Feature request & issues

If anything is missing, unclear, or simply not working, open an issue on the repository.

What is left to do?

• Better background generation
• Better handwritten text generation
• More customization parameters (mostly regarding background)
• Implement --include_symbols
• Implement --include_numbers