Nearly 99% accuracy

[Des Bw]

I am getting nearly 99% accuracy by training from the top layer of the network. I am training using synthetic data; and the evaluation is done the same type of data. 

But, the result is not extending to actually scanned documents. On the scanned documents, I am getting lower accuracy, unfortunately. 

 So, it looks like training using synthetic data is actually not that productive approach.