> 2) While building the ocropus project I got: > WARNING: scons not supported on platforms other than Ubuntu 9.10 > as I reported previously. Nobody expressed an opinion about this and > things looked OK, so I pressed on.
Yes; the meaning of that is that we simply don't have the resources to test on more than one platform. That identifies the current platform we're building and testing on. > 4) When I tried to run ocropus I discovered I needed to install > python-matplotlib. There might be other python dependencies not listed > in ubuntu-install; I already have numpy installed, for example. Yes, not all the Python dependencies are listed yet. That's on our cleanup list. > 5) The results of building all the modules were a little weird: some of > the files ended up installed to /usr/local, while others went to /usr. > In particular, python modules ended up > in /usr/lib/python2.5/site-packages rather > than /usr/local/lib/python2.5/site-packages. OCRopus data files go into /usr/local/..., Python files go wherever setup.py installs them. > 6) > $ ocropus-pages -h > Traceback (most recent call last): > File "/usr/bin/ocropus-pages", line 13, in <module> > import ocropy > File "/usr/lib/python2.5/site-packages/ocropy/__init__.py", line 4, in > <module> > from ocropus import * > File "/usr/lib/python2.5/site-packages/ocropus.py", line 10, in > <module> > import _ocropus > ImportError: libocropus.so: cannot open shared object file: No such file > or directory Check whether there is a "/usr/local/lib/libocropus.so". If not, your build failed somehow; you need to figure out why it didn't get installed. If there is, you need to add /usr/local/lib to your LD_LIBRARY_PATH so that Python can find it. > but ocropus page foo.png ran. The former is in /usr/bin, the latter > in /usr/local/bin. /usr/bin/ocropus sounds like it was installed from a package rather than from the sources. The ocropus command itself is pure C++ and doesn't use Python. Try "dpkg -S /usr/bin/ocropus". If that yields a result, you are running the packaged version. > 7) I tried ocropus page on a receipt and a medical insurance statement. > It didn't do too well. I realize these are outside of ocropus's target > useage. Unfortunately, they are the primary documents I'm interested > in. The models that OCRopus ships with have been trained on a large database of printed materials, but it turns out that that database doesn't contain a large variety of fonts, so it doesn't work well on many other materials. We're currently making a database of documents in many more printed fonts and training OCRopus on that. With that, you should get better performance on a wider variety of materials. 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.
