Testing Code File

[HiteshG]

Which code file is required to directly test images dataset.

[Breta01]

For evaluation there is notebooks/OCR.ipynb it should cover whole process, just change the location of image. RIght now it supports only one type of model. For different methods look into the notebooks/ocr_evaluator.ipynb. Along with that you have to download pretrained models, for more information look into the models/ directory. Keep in mind that the current models haven't been trained on the new datasets so they can perform quite poorly on some types of handwriting.

[HiteshG]

in notebooks/ocr_evaluuator.ipynb what is " images, labels = load_words_data('../data/sets/test.csv', is_csv=True) " There is no set folder in data. Do i have to create test.csv for my test input ? If yes then how ?

[HiteshG]

I want to test a single image of the page for Bi-RNN and CNN, so how to use ocr_evaluator.ipynb.

[Breta01]

Ok, I admit that the documentation is bad (there isn't any documentation). However I get notified about your comments and there's no need to send me two emails on top of that. This is open source projects and I have a lot of other stuff to do. Be patient or donate the project.

If you want to evaluate a single image, look into the OCR.ipynb, there's no need to create test.csv. Then replace the code for recognition of words with the code for recognition of words using CTC from ocr_evaluator.ipynb.

[HiteshG]

Yeah sure, I will contribute to the documentation :)

[Breta01]

Thanks, that would be great.