curAsciiDir = os.path.join(asciiPrefix, l.split('-')[0], l[0:7])
curAsciiFilePath = os.path.join(curAsciiDir, l + ".txt")
[curTargetString, curWordTargetString,
curCharSet] = getTargetString(curAsciiFilePath)
targetStrings.extend(curTargetString)
wordTargetStrings.extend(curWordTargetString)
# print len(curTargetString)
# print curCharSet
charSet = charSet.union(curCharSet)
# for i in range(len(curTargetString)):
# print curWordTargetString[i]
# print curTargetString[i]
xmlNames = getXmlNames(curXmlDir, l)
assert len(curTargetString) == len(xmlNames)
for xmlName in xmlNames:
seqTags.append(xmlName)
xmlFilePath = os.path.join(curXmlDir, xmlName)
curLineStroke = getLineStroke(xmlFilePath)
# print len(curLine)
inputs.extend(curLineStroke)
seqLengths.append(len(curLineStroke))
seqDims.append([len(curLineStroke)])
inputsArr = np.array(inputs)
inputMeans = np.mean(inputsArr, 0)
inputStds = np.std(inputsArr, 0)
inputsArr[:, :-1] = (inputsArr[:, :-1] - inputMeans[:-1]) / inputStds[:-1]
inputs = inputsArr.tolist()
for l in trainFileList:
l = l.strip()
print "loading file ", l
#print l[0:7]
curXmlDir = os.path.join(xmlPrefix, l.split('-')[0], l[0:7])
curAsciiDir = os.path.join(asciiPrefix, l.split('-')[0], l[0:7])
curAsciiFilePath = os.path.join(curAsciiDir, l + ".txt")
[curTargetString, curWordTargetString, curCharSet] = getTargetStringCompress(curAsciiFilePath)
targetStrings.extend(curTargetString)
wordTargetStrings.extend(curWordTargetString)
#print len(curTargetString)
#print curCharSet
charSet = charSet.union(curCharSet)
xmlNames = getXmlNames(curXmlDir, l)
for xmlName in xmlNames:
seqTags.append(xmlName)
xmlFilePath = os.path.join(curXmlDir, xmlName)
curLineStroke = getLineStrokeOffset(xmlFilePath)
#print len(curLine)
inputs.append([0.0] * 3)
inputs.extend(curLineStroke)
inputs = inputs[:-1]
seqLengths.append(len(curLineStroke))
seqDims.append([len(curLineStroke)])
for coord in curLineStroke:
targetPatterns.append(coord[:-1]);
targetClasses.append([coord[-1]]);
inputsArr = np.array(inputs)