Merge branch 'master' into segment-improve-sep-line-detection

Author: bertsky

README.md

@@ -11,7 +11,7 @@ Content:
          * [ocrd-cis-postcorrect](#ocrd-cis-postcorrect)
          * [ocrd-cis-align](#ocrd-cis-align)
          * [ocrd-cis-data](#ocrd-cis-data)
-         * [Trainining](#trainining)
+         * [Training](#training)
          * [ocrd-cis-ocropy-train](#ocrd-cis-ocropy-train)
          * [ocrd-cis-ocropy-clip](#ocrd-cis-ocropy-clip)
          * [ocrd-cis-ocropy-resegment](#ocrd-cis-ocropy-resegment)
@@ -74,11 +74,11 @@ It is possible (and recommended) to install `ocrd_cis` in a custom user director
 ```
 
 ## Profiler
-The post correction is dependent on the language
-[profiler](https://github.com/cisocrgroup/Profiler) and its laguage
-configurations to generate corrections for suspicious words.  In order
-to use the post correction a profiler with according language
-configruations have to be present on the system.  You can refer to our
+The post-correction is dependent on the language
+[profiler](https://github.com/cisocrgroup/Profiler) and its language
+configurations to generate corrections for suspicious words. In order
+to use the post-correction, a profiler and according language
+configurations have to be present on the system. You can refer to our
 [manuals](https://github.com/cisocrgroup/Resources/tree/master/manuals)
 and our [lexical
 resources](https://github.com/cisocrgroup/Resources/tree/master/lexica)
@@ -162,7 +162,7 @@ jar library, the pre-trained post correction model, the path to the
 default 3-grams language model file or the default training
 configuration file.  This tool does not follow the OCR-D conventions.
 
-### Trainining
+### Training
 There is no dedicated training script provided. Models are trained
 using the java implementation directly (check out the [training test
 script](tests/run_training_test.bash) for an example).  Training a
@@ -480,7 +480,7 @@ installed in order to run any tests).
 # Miscellaneous
 ## OCR-D workspace
 
-* Create a new (empty) workspace: `ocrd workspace init workspace-dir`
+* Create a new (empty) workspace: `ocrd workspace -d workspace-dir init`
 * cd into `workspace-dir`
 * Add new file to workspace: `ocrd workspace add file -G group -i id
   -m mimetype -g pageId`

ocrd_cis/aio/aio.py

@@ -161,7 +161,7 @@ def addtoworkspace(wsdir, gtdir):
 
     _, _, files = os.walk(wsdir).__next__()
     if 'mets.xml' not in files:
-        initcmd = 'ocrd workspace init {}'.format(wsdir)
+        initcmd = 'ocrd workspace -d {} init'.format(wsdir)
         subprocess_cmd(initcmd)
 
 

ocrd_cis/ocropy/ocrolib/common.py

@@ -19,8 +19,6 @@
                                 OcropusException)
 from numpy import (amax, amin, array, bitwise_and, clip, dtype, mean, minimum,
                    nan, sin, sqrt, zeros, unique, fromstring)
-from pylab import (clf, cm, ginput, gray, imshow, ion, subplot,
-                   where, xticks, yticks, title, xlabel, ylabel)
 from scipy.ndimage import morphology, measurements
 import PIL
 
@@ -802,6 +800,7 @@ def binarize_range(image,dtype='B',threshold=0.5):
 
 def plotgrid(data,d=10,shape=(30,30)):
     """Plot a list of images on a grid."""
+    from matplotlib.pyplot import ion, gray, clf, subplot, imshow, ginput
     ion()
     gray()
     clf()
@@ -813,11 +812,13 @@ def plotgrid(data,d=10,shape=(30,30)):
     ginput(1,timeout=0.1)
 
 def showrgb(r,g=None,b=None):
+    from matplotlib.pyplot import imshow
     if g is None: g = r
     if b is None: b = r
     imshow(array([r,g,b]).transpose([1,2,0]))
 
 def showgrid(l,cols=None,n=400,titles=None,xlabels=None,ylabels=None,**kw):
+    from matplotlib.pyplot import cm, xticks, yticks, subplot, imshow, title, xlabel, ylabel
     if "cmap" not in kw: kw["cmap"] = cm.gray
     if "interpolation" not in kw: kw["interpolation"] = "nearest"
     n = minimum(n,len(l))

ocrd_cis/ocropy/ocrolib/ligatures.py

@@ -5,7 +5,7 @@
 ################################################################
 
 import re
-from pylab import uint32
+from numpy import uint32
 
 ### These aren't formal ligatures, they are character pairs
 ### that are frequently touching in Latin script documents.

ocrd_cis/ocropy/ocrolib/lineest.py

@@ -3,7 +3,6 @@
 import os
 
 import numpy as np
-import matplotlib.pyplot as plt
 from scipy.ndimage import interpolation,filters,measurements
 
 def scale_to_h(img,target_height,order=1,dtype=np.dtype('f'),cval=0):
@@ -71,6 +70,7 @@ def measure(self,line):
         self.mad = np.mean(deltas[line!=0])
         self.r = int(1+self.range*self.mad)
         if self.debug:
+            import matplotlib.pyplot as plt
             plt.figure("center")
             plt.imshow(line,cmap=plt.cm.gray)
             plt.plot(self.center)

ocrd_cis/ocropy/ocrolib/lstm.py

@@ -30,7 +30,6 @@
 import unicodedata, sys
 
 import numpy as np
-import matplotlib.pyplot as plt
 from scipy.ndimage import measurements,filters
 
 from . import common as ocrolib
@@ -806,6 +805,7 @@ def ctc_align_targets(outputs,targets,threshold=100.0,verbose=0,debug=0,lo=1e-5)
     lmatch = np.log(match)
 
     if debug:
+        import matplotlib.pyplot as plt
         plt.figure("ctcalign"); plt.clf();
         plt.subplot(411); plt.imshow(outputs.T,interpolation='nearest',cmap=plt.cm.hot)
         plt.subplot(412); plt.imshow(lmatch.T,interpolation='nearest',cmap=plt.cm.hot)

ocrd_cis/ocropy/ocrolib/morph.py

@@ -5,7 +5,6 @@
 
 
 from numpy import *
-import pylab
 #from scipy.ndimage import morphology,measurements,filters
 from scipy.ndimage import measurements
 from scipy.ndimage.interpolation import shift
@@ -154,7 +153,8 @@ def rg_closing(image,size,origin=0):
 
 @checks(SEGMENTATION)
 def showlabels(x,n=7):
-    pylab.imshow(where(x>0,x%n+1,0),cmap=pylab.cm.gist_stern)
+    import matplotlib.pyplot as plt
+    plt.imshow(where(x>0,x%n+1,0),cmap=plt.cm.gist_stern)
 
 @checks(ABINARY2)
 def find_contours(image):

ocrd_cis/ocropy/ocrolib/psegutils.py

@@ -1,8 +1,6 @@
 from __future__ import print_function
 
 import numpy as np
-import matplotlib.pyplot as plt
-import matplotlib.patches as mpatches
 from scipy.ndimage import filters,interpolation
 
 from .toplevel import *
@@ -131,6 +129,7 @@ def separates(w,u,v):
         if w[1].start<u[1].stop and w[1].stop>v[1].start: return 1
         return 0
     if highlight is not None:
+        import matplotlib.pyplot as plt
         plt.clf()
         plt.title("highlight")
         plt.imshow(lines)
@@ -178,6 +177,8 @@ def find(condition):
 def show_lines(image,lines,lsort):
     """Overlays the computed lines on top of the image, for debugging
     purposes."""
+    import matplotlib.pyplot as plt
+    import matplotlib.patches as mpatches
     ys,xs = [],[]
     plt.clf()
     plt.cla()
@@ -197,12 +198,14 @@ def show_lines(image,lines,lsort):
 
 @obsolete
 def read_gray(fname):
+    import matplotlib.pyplot as plt
     image = plt.imread(fname)
     if image.ndim==3: image = np.mean(image,2)
     return image
 
 @obsolete
 def read_binary(fname):
+    import matplotlib.pyplot as plt
     image = plt.imread(fname)
     if image.ndim==3: image = np.mean(image,2)
     image -= np.amin(image)
@@ -214,6 +217,7 @@ def read_binary(fname):
 @obsolete
 def rgbshow(r,g,b=None,gn=1,cn=0,ab=0,**kw):
     """Small function to display 2 or 3 images as RGB channels."""
+    import matplotlib.pyplot as plt
     if b is None: b = np.zeros(r.shape)
     combo = np.transpose(np.array([r,g,b]),axes=[1,2,0])
     if cn:

setup.py

@@ -46,10 +46,12 @@
         'shapely>=1.7.1',
         'scikit-image',
         'opencv-python-headless',
-        'matplotlib>3.0.0',
         'python-Levenshtein',
         'calamari_ocr == 0.3.5'
     ],
+    extras_require={
+        'debug': ['matplotlib>3.0.0'],
+    },
     package_data={
         '': ['*.json', '*.yml', '*.yaml', '*.csv.gz', '*.jar', '*.zip'],
     },