def getSegmentationAccuracy(pathArray):
dataX = []
dataY = []
j = 0
for path in pathArray:
for f in os.listdir(path):
j+=1
if f[-5:] != "inkml": continue
# print f
traceList, symbolsList = gettrace.parseINKMLFile(path + f)
segmentIndices = segment.segmentSymbols(traceList)
for label, elem in symbolsList:
# if label in ['=', 'i', 'j','\\leq', '\\log', '\\sin', '\\cos', '\\lim', '\\geq', '\\righarrow', '\\div']: continue
# Correctly classified
if elem in segmentIndices:
strokes = [traceList[i] for i in elem]
pixels = itp.inkml_to_pixels(strokes)
chain = list(itertools.chain(*pixels))
chain.append(len(strokes))
dataX.append(chain)
dataY.append(label)
# cr.display(pixels)
print j
f = file("segmented_data_18", "w")
pickle.dump((dataX, dataY), f)
def makeSpecialSymbol(prev_trace, trace): global svm global labels pixels = itp.inkml_to_pixels([prev_trace, trace]) chain = list(itertools.chain(*pixels)) chain.append(2) probabilities = svm.predict_proba(chain) probMax = np.amax(probabilities) indexOfMax = np.argmax(probabilities) if labels[indexOfMax] in ['x', 'k', '\\geq', 'i', 'j', '='] and probMax > 0.7: return True return False
def recognize():
global svm
print request.form['info']
strokes = json.loads(request.form['info'])
symbolsIndices = segment.segmentSymbols(strokes)
returnstr = ""
for elem in symbolsIndices:
s = [strokes[i] for i in elem]
pixels = inkml_to_pixels.inkml_to_pixels(s)
chain = list(itertools.chain(*pixels))
chain.append(len(s))
prediction = svm.predict(chain)[0]
returnstr += prediction
return returnstr
def getSegmentationAccuracy(pathArray):
f = open('svm18px', "r")
svm = pickle.load(f)
j = 0
total = 0
total_correct = 0
off_by_one = 0
off_by_two = 0
for path in pathArray:
for f in os.listdir(path):
j+=1
if f[-5:] != "inkml": continue
# print f
traceList, symbolsList = gettrace.parseINKMLFile(path + f)
segmentIndices = segment.segmentSymbols(traceList)
correct = 0
for label, elem in symbolsList:
# Correctly classified
if elem in segmentIndices:
strokes = [traceList[i] for i in elem]
pixels = itp.inkml_to_pixels(strokes)
chain = list(itertools.chain(*pixels))
chain.append(len(strokes))
if svm.predict(chain)[0] == label:
correct +=1
if correct == len(symbolsList):
total_correct +=1
if correct + 1 >= len(symbolsList):
off_by_one +=1
if correct + 2 >= len(symbolsList):
off_by_two +=1
# cr.display(pixels)
print j
print total_correct
print off_by_one
print off_by_two
print float(total_correct)/836
print float(off_by_one)/836
print float(off_by_two)/836
def recognize_hmm():
global hmm_instance
strokes = json.loads(request.form['info'])
symbolsIndices = segment.segmentSymbols(strokes)
returnstr = ""
equation = []
for elem in symbolsIndices:
s = [strokes[i] for i in elem]
pixels = inkml_to_pixels.inkml_to_pixels(s)
chain = list(itertools.chain(*pixels))
chain.append(len(s))
equation.append((chain,))
prediction = hmm_instance.compute_best_sequence(equation)
for symbol in prediction:
returnstr += symbol
return returnstr
def getData(pathArray):
dataX = []
dataY = []
from time import time
start = time()
for path in pathArray:
for file in os.listdir(path):
if file[-5:] != "inkml": continue
traceList, symbolsList = gettrace.parseINKMLFile(path + file)
for label, indices in symbolsList:
strokes = [traceList[elem] for elem in indices]
pixels = itp.inkml_to_pixels(strokes)
chain = list(itertools.chain(*pixels))
chain.append(len(strokes))
dataX.append(chain)
dataY.append(label)
# itp.display(pixels)
print time()-start
return dataX, dataY
def getData(pathArray):
data = []
from time import time
start = time()
for path in pathArray:
for file in os.listdir(path):
if file[-5:] != "inkml": continue
traceList, symbolsList = gettrace.parseINKMLFile(path + file)
equationDict = {}
for label, indices in symbolsList:
strokes = [traceList[elem] for elem in indices]
pixels = itp.inkml_to_pixels(strokes)
chain = list(itertools.chain(*pixels))
chain.append(len(strokes))
equationDict[indices[0]] = (chain, label)
equationList = []
for key in sorted(equationDict):
equationList.append(equationDict[key])
data.append(equationList)
# itp.display(pixels)
print time()-start
return data
