def get_linenormalizer():
global network
global lnorm
# load the network used for classification
try:
network = ocrolib.load_object(args['model'], verbose=1)
for x in network.walk():
x.postLoad()
for x in network.walk():
if isinstance(x, lstm.LSTM):
x.allocate(5000)
except FileNotFound:
print_error("")
print_error("Cannot find OCR model file:" + args['model'])
print_error("Download a model and put it into:" +
ocrolib.default.modeldir)
print_error("(Or override the location with OCROPUS_DATA.)")
print_error("")
sys.exit(1)
# get the line normalizer from the loaded network, or optionally
# let the user override it (this is not very useful)
lnorm = getattr(network, "lnorm", None)
if args['height'] > 0:
lnorm.setHeight(args['height'])
def initRnnModel(model_name="en-default.pyrnn"):
"""
This function load the neural network model from slapos backend
and initialise it.
----------------------------
@args:
- model_name: string, default: en-default.pyrnn
Id of the object in the filesystem that contain the rnn model
@return:
- network: lstm object
Represent the trained neural net
- lnorm: method from lstm object
Represent the size of the lstm object. Is used to scale the objects
to recognize from original size to the average network object.
----------------------------
LSTM meaning: https://en.wikipedia.org/wiki/Long_short-term_memory
lnorm is extracted for clarity. This function initialize the neural net after
loading.
"""
network = ocrolib.load_object(model_name)
for node in network.walk():
if isinstance(node, ocrolib.lstm.LSTM):
node.allocate(5000)
lnorm = getattr(network, "lnorm", None)
return network, lnorm
def load_lstm(fname):
if args.clstm:
network = lstm.SeqRecognizer(args.height,args.hiddensize,
codec=codec,
normalize=lstm.normalize_nfkc)
import clstm
mylstm = clstm.make_BIDILSTM()
mylstm.init(network.No,args.hiddensize,network.Ni)
mylstm.load(fname)
network.lstm = clstm.CNetwork(mylstm)
return network
else:
network = ocrolib.load_object(last_save)
network.upgrade()
for x in network.walk(): x.postLoad()
return network
parser.add_argument("files",
nargs="+",
help="input files; glob and @ expansion performed")
args = parser.parse_args()
files = args.files
str1 = files[0]
fname_out = str1[0:str1.rfind('-')] + '-error'
f = open(fname_out + '.txt', 'w')
err = []
count = 0
m_name = []
for model in files:
network = ocrolib.load_object(model)
err.append(network.error)
m_name.append(model)
count += 1
data = {}
data['minimum'] = {
'name': m_name[err.index(min(err))],
'error': min(list(err))
}
data['last'] = {'name': m_name[-1], 'error': list(err)[-1]}
json.dump(data, f)
f.close()
def ocr(image_path, segmentation_path, output_path, model_path):
"""
Scan a single image with ocropus.
Reads a single image file from ```imagepath``` and writes the recognized
text as a TEI document into output_path.
Args:
image_path (unicode): Path of the input file
segmentation_path (unicode): Path of the segmentation XML file.
output_path (unicode): Path of the output file
model_path (unicode): Path of the recognition model. Must be a pyrnn.gz
pickle dump interoperable with ocropus-rpred.
Returns:
(unicode): A string of the output file that is actually written. As
Ocropus rewrites output file paths without notice it may be
different from the ```outputfilepath``` argument.
Raises:
NidabaOcropusException: Ocropus somehow failed. The error output is
contained in the message but as it is de facto
unusable as a library it's impossible to deduct
the nature of the problem.
"""
try:
logger.debug('Loading pyrnn from {}'.format(model_path))
network = ocrolib.load_object(model_path, verbose=0)
lnorm = getattr(network, "lnorm")
except Exception as e:
raise NidabaOcropusException('Something somewhere broke: ' + e.msg)
im = Image.open(image_path)
logger.debug('Loading TEI segmentation {}'.format(segmentation_path))
tei = TEIFacsimile()
with open(segmentation_path, 'r') as seg_fp:
tei.read(seg_fp)
logger.debug('Clearing out word/grapheme boxes')
# ocropus is a line recognizer
tei.clear_graphemes()
tei.clear_segments()
# add and scope new responsibility statement
tei.add_respstmt('ocropus', 'character recognition')
for box in tei.lines:
logger.debug('Recognizing line {}'.format(box[4]))
ib = tuple(int(x) for x in box[:-2])
line = ocrolib.pil2array(im.crop(ib))
temp = np.amax(line) - line
temp = temp * 1.0 / np.amax(temp)
lnorm.measure(temp)
line = lnorm.normalize(line, cval=np.amax(line))
if line.ndim == 3:
np.mean(line, 2)
line = ocrolib.lstm.prepare_line(line, 16)
pred = network.predictString(line)
pred = ocrolib.normalize_text(pred)
logger.debug('Scoping line {}'.format(box[4]))
tei.scope_line(box[4])
logger.debug('Adding graphemes: {}'.format(pred))
tei.add_graphemes(pred)
with open(output_path, 'wb') as fp:
logger.debug('Writing TEI to {}'.format(fp.abs_path))
tei.write(fp)
return output_path
def ocr(image_path, segmentation_path, output_path, model_path):
"""
Scan a single image with ocropus.
Reads a single image file from ```imagepath``` and writes the recognized
text as a TEI document into output_path.
Args:
image_path (unicode): Path of the input file
segmentation_path (unicode): Path of the segmentation XML file.
output_path (unicode): Path of the output file
model_path (unicode): Path of the recognition model. Must be a pyrnn.gz
pickle dump interoperable with ocropus-rpred.
Returns:
(unicode): A string of the output file that is actually written. As
Ocropus rewrites output file paths without notice it may be
different from the ```outputfilepath``` argument.
Raises:
NidabaOcropusException: Ocropus somehow failed. The error output is
contained in the message but as it is de facto
unusable as a library it's impossible to deduct
the nature of the problem.
"""
try:
logger.debug('Loading pyrnn from {}'.format(model_path))
network = ocrolib.load_object(model_path, verbose=0)
lnorm = getattr(network, "lnorm")
except Exception as e:
raise NidabaOcropusException('Something somewhere broke: ' + e.msg)
im = Image.open(image_path)
logger.debug('Loading TEI segmentation {}'.format(segmentation_path))
tei = TEIFacsimile()
with open(segmentation_path, 'r') as seg_fp:
tei.read(seg_fp)
logger.debug('Clearing out word/grapheme boxes')
# ocropus is a line recognizer
tei.clear_graphemes()
tei.clear_segments()
# add and scope new responsibility statement
tei.add_respstmt('ocropus', 'character recognition')
for box in tei.lines:
logger.debug('Recognizing line {}'.format(box[4]))
ib = tuple(int(x) for x in box[:-2])
line = ocrolib.pil2array(im.crop(ib))
temp = np.amax(line) - line
temp = temp * 1.0 / np.amax(temp)
lnorm.measure(temp)
line = lnorm.normalize(line, cval=np.amax(line))
if line.ndim == 3:
np.mean(line, 2)
line = ocrolib.lstm.prepare_line(line, 16)
pred = network.predictString(line)
pred = ocrolib.normalize_text(pred)
logger.debug('Scoping line {}'.format(box[4]))
tei.scope_line(box[4])
logger.debug('Adding graphemes: {}'.format(pred))
tei.add_graphemes(pred)
with open(output_path, 'wb') as fp:
logger.debug('Writing TEI to {}'.format(fp.abs_path))
tei.write(fp)
return output_path
print_info("")
print_info("#" * 10, (" ".join(sys.argv))[:60])
print_info("")
inputs = ocrolib.glob_all(args.files)
if not args.quiet: print_info("#inputs" + str(len(inputs)))
# disable parallelism when anything is being displayed
if args.show >= 0 or args.save is not None:
args.parallel = 1
# load the network used for classification
try:
network = ocrolib.load_object(args.model, verbose=1)
for x in network.walk(): x.postLoad()
for x in network.walk():
if isinstance(x, lstm.LSTM):
x.allocate(5000)
except ocrolib.FileNotFound:
print_error("")
print_error("Cannot find OCR model file:" + args.model)
print_error("Download a model and put it into:" + ocrolib.default.modeldir)
print_error("(Or override the location with OCROPUS_DATA.)")
print_error("")
sys.exit(1)
# get the line normalizer from the loaded network, or optionally
# let the user override it (this is not very useful)
import h5py
import numpy
import json
import argparse
import ocrolib
parser = argparse.ArgumentParser("convert pyrnn to clstm")
parser.add_argument("file")
parser.add_argument("-o","--out",default="en-default.hdf5",
help="Filename to export the parameters")
args = parser.parse_args()
recognizer = ocrolib.load_object(args.file)
# TODO: load gzipped things
# recognizer = pickle.load(open('en-default.pyrnn'))
parallel, softmax = recognizer.lstm.nets
fwdnet, revnet = parallel.nets
nf = h5py.File(args.out, "w")
for w in "WGI WGF WGO WCI".split():
print getattr(fwdnet, w).shape
dset = nf.create_dataset(".bidilstm.0.parallel.0.lstm." + w, getattr(fwdnet, w).shape, dtype='f')
dset[...] = getattr(fwdnet, w)
dset = nf.create_dataset(".bidilstm.0.parallel.1.reversed.0.lstm." + w, getattr(revnet.net, w).shape, dtype='f')
dset[...] = getattr(revnet.net, w)