Issues with Models?

[Adeel-Intizar]

Hello, Thank you for your great work
The model does not provide accurate results with different backgrounds
and what should be the resolution of input images?

[HassamChundrigar]

The model is trained for document images, which have white background and black color text, However training data have noisy images too.
the input dimentions are 64x1024
resized_image = tf.image.resize_image_with_pad(image, 64, 1024).eval(session = sess)

[rizMehdi]

Hey @HassamChundrigar Hassam. Thank you for your awesome work. i am using noisy images, but i convert them to black and white and resize them to 64,1024, but i dont get accurate results. I think i need to re-train the model to get better results and the urdu font in my images is slightly bold then the one you used. Any directions or tip for training the model?