General questions

[ghost]

@ankanbhunia Thank you for your work.

• What is the license for this project?
• What are the commands used for training?
• What are the commands used to binarize an image?
• Which version of OpenCV is required?
• How did you convert the images from those datasets into patches of size 256 X 256?

[ankanbhunia]

@MrOCR Thanks for your interest in our work.

We use OpenCV version 3.4.5. You can start training the model just by running the train.py. You can input the dataset path and adjust the other parameters (Batch_size, No_of_epochs, etc.) by editing the params dictionary in that file. We follow the same procedure to extract the patches as described in DSN-Binarization .You can download these extracted patches from here. After training the model, If you want to get the binarized output from testing images, you have to use the function Noise_remover_network. The full testing code is not provided in this repository.
And to answer your first question, you can use our code for non-commercial research purposes.

[zokai]

Hi @ankanbhunia, Is it possible to run your code on a Windows 10 environment? Tensorflow for Python 2.7 only available for Linux and Mac. Is it possible to modify your code to work on Python 3.6?

Thank you

[ankanbhunia]

I think you can run the same code on Python 3.6 as well.

[unik00]

You can use my testing code. https://github.com/unik00/AdverseBiNet

[pra-dan]

Hey @unik00 Can you provide a more detailed way to test ? Maybe a Google Collab notebook ? Thanks.

[SB2020-eye]

Like others before me (above), I would be so grateful for brief, simple guidance about how to use the code/repository to get nice results like @unik00 posted above.

Thanks.

[jrdyhwh]

Is there a way to use this code in tf 2.0?