>
>
>                           thank you for your informative reply. We are
> again test the performance of bpnet where we train the isolated characters
> with single color and trying to recognize the test images. We are
> continuously experimenting the parameters of bpnet. However we are yet to
> get 100% accuracy even to recognize the training image. Can you please
> suggest us whether we are missing anything which is eventually effecting the
> recognition?


Well, as I have said before, you shouldn't expect high performance OCR from
bpnet; bpnet is a simple neural network implementation meant originally
handwritten character recognition; it's in OCRopus so that we can experiment
with training and APIs, not as a high performance OCR solution.

In addition, many classifiers just don't behave the way you expect; they are
not guaranteed to recognize every character in the training set correctly,
and training them on a small set of training examples makes this effect
worse.

We're currently working on a new set of character recognizers.  The goal is
to deal with a number of issues: clean up the code, give competitive error
rates, support larger character sets, and automate training so that you
don't have to worry about as many issues.  We have implemented those methods
in Python and are now porting them to C++.

Porting the code to C++ is, unfortunately, a lot of work, not because of the
learning algorithms themselves (that's easy), but because we need to write a
lot of code for pickling, factory methods, error checking, etc.

So, it's good for you to get started on training with bpnet, because
training with the new recognizers will work pretty much the same way.  But
until the new recognizers are implemented, don't expect really good
performance from bpnet.

A second effect is that if you use OCRopus with bpnet, there is no effective
language model (dictionary) yet.  The next release will have language models
comparable to those found in other OCR systems, with the option of training
even better language models yourself.

Tom

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"ocropus" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/ocropus?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to