That's probably related to the first problem: there is no training data at all, apparently because there is something wrong with the segmentations on your machines.
Tom On Tuesday, August 21, 2012 11:41:41 PM UTC+2, stinger wrote: > > Also, when running the run-uw3-500 script, it fails at ocopus-tsplit with > the following error - as if the book.h5 file is corrupt? Any help is > appreciated: > > ocropus-tsplit -d book.h5 -o book.tsplit --maxsplit 100 > loading dataset > got 0 samples out of 0 > # classes 0 > most common ... > starting training > pcakmeans 0 k 0 d 0.95 > Traceback (most recent call last): > File "/usr/local/bin/ocropus-tsplit", line 137, in <module> > sc.fit(patches) > File "/usr/local/lib/python2.7/dist-packages/ocrolib/patrec.py", line > 355, in fit > self.splitter.fit(data) > File "/usr/local/lib/python2.7/dist-packages/ocrolib/patrec.py", line > 289, in fit > maxiter=self.maxiter,npk=self.npk,verbose=self.verbose) > File "/usr/local/lib/python2.7/dist-packages/ocrolib/toplevel.py", line > 186, in argument_checks > raise CheckError(e.message,*e.args,var=var,fun=f) > ocrolib.toplevel.CheckError: > CheckError for argument 'data' in call to function: '<function pca_kmeans > at 0x37367d0>' > > > > On Tuesday, August 21, 2012 4:17:57 PM UTC-5, stinger wrote: >> >> I've run through the installation process, and am trying to run the >> run-test script. I keep getting the error "expected a segmentation with >> white background" when running ocropus-ngraphs (see output below) - it's >> failing because the max of 255 is being checked against an expected value >> of 0xffffff. However, if I modify the code, and change 0xffffff to 0xff >> the run-test script works. Not sure if this is a bug? >> >> + true >> + true language model application >> + true >> + ocropus-ngraphs 'temp/????/??????.lattice' >> loading /usr/local/share/ocropus/en-mixed-4.ngraphs >> processing 92 files >> temp/0001/010001.lattice =NGRAPHS= 21.29 BOOK REVIEIP >> Traceback (most recent call last): >> File "/usr/local/bin/ocropus-ngraphs", line 294, in <module> >> rseg = ocrolib.read_line_segmentation(rname) >> File "/usr/local/lib/python2.7/dist-packages/ocrolib/toplevel.py", line >> 194, in argument_checks >> result = f(*args,**kw) >> File "/usr/local/lib/python2.7/dist-packages/ocrolib/common.py", line >> 207, in read_line_segmentation >> result = make_seg_black(image) >> File "/usr/local/lib/python2.7/dist-packages/ocrolib/toplevel.py", line >> 190, in argument_checks >> raise e >> ocrolib.toplevel.CheckError: >> CheckError for argument 'image' in call to function: '<function >> make_seg_black at 0x3810aa0>' >> <ndarray-13b67fd0 (60, 583) int32 [1,255]> of type <type >> 'numpy.ndarray'>: expected a segmentation with white background >> >> >> >> >> >> >> On Friday, August 17, 2012 9:31:35 PM UTC-5, Tom wrote: >>> >>> OCRopus 0.6pre1 has been released. It features much simpler >>> installation, fewer dependencies, and improved recognition rates. This is >>> the first all-Python release. Please follow the instructions on >>> http://www.ocropus.org/ (installation is really just a couple of simple >>> steps). >>> >>> There are three scripts you should run after installation: >>> >>> (1) "run-test" runs a simple recognition test >>> (2) "run-box-training" (in fraktur-boxes) trains a Fraktur recognizer >>> from Tesseract-style box files >>> (3) "run-uw3-500" (in uw3-500) shows how training works on line-by-line >>> transcribed data >>> >>> 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]. To view this discussion on the web visit https://groups.google.com/d/msg/ocropus/-/-leH6_WupAYJ. For more options, visit https://groups.google.com/groups/opt_out.
