def createGTUnique(classesDict, length, gtList):
"""
Create ground truth array where only one label is allowed per point
@param classesDict:
@param length: length of the final array (=length of prediction array)
@param gtList:
@return:
"""
y_GT = np.empty([length])
y_GT.fill(-1) #-1 corresponds to no label given
classesNotTrained = []
for i in range(len(gtList)):
""" Fill array from start to end of each ground truth label with the correct label: """
if gtList[i][2] == "start":
tmpContext = gtList[i][1]
start = getIndex(float(gtList[i][0]))
# Find the end time of this context:
for j in range(i,len(gtList)):
if ((gtList[j][1] == tmpContext) and (gtList[j][2] == "end")):
end = getIndex(float(gtList[j][0]))
if end >= y_GT.shape[0]:
end = y_GT.shape[0] - 1
""" Fill ground truth array, and check if our classifier was
trained with all labels of the test file, if not give warning: """
if gtList[i][1] not in classesDict.keys():
classesNotTrained.append(gtList[i][1])
y_GT[start:end+1].fill(-1)
else:
y_GT[start:end+1].fill(classesDict[tmpContext])
break
if classesNotTrained:
for el in set(classesNotTrained):
print("The classifier wasn't trained with class '" +
el + "'. It will not be considered for testing.")
return y_GT
def createGTMulti(classesDict, length, gtList):
"""
Create ground truth array that allows multiple labels per point
@param classesDict:
@param length: length of the final array (=length of prediction array)
@param gtList:
@return:
"""
""" Create array containing label for sample point: """
n_maxLabels = 5 #maximum number of labels that can be assign to one point
y_GT = np.empty([length,n_maxLabels])
y_GT.fill(-1) #-1 corresponds to no label given
classesNotTrained = []
for i in range(len(gtList)):
""" Fill array from start to end of each ground truth label with the correct label: """
gtList[i][2]
if gtList[i][2] == "start":
tmpContext = gtList[i][1]
start = getIndex(float(gtList[i][0]))
# Find the end time of this context:
for j in range(i,len(gtList)):
if ((gtList[j][1] == tmpContext) and (gtList[j][2] == "end")):
end = getIndex(float(gtList[j][0]))
if end >= y_GT.shape[0]:
end = y_GT.shape[0] - 1
""" Fill ground truth array, and check if our classifier was
trained with all labels of the test file, if not give warning: """
if (gtList[i][1] not in classesDict.keys()):
classesNotTrained.append(gtList[i][1])
else:
# Check if we can write into the first column of the y_GT array:
if ((len(np.unique(y_GT[start:end+1,0])) == 1) and
(np.unique(y_GT[start:end+1,0])[0] == -1)):
y_GT[start:end+1,0].fill(classesDict[gtList[i][1]])
# Check if we can write into the second column of the y_GT array:
elif ((len(np.unique(y_GT[start:end+1,1])) == 1) and
(np.unique(y_GT[start:end+1,1])[0] == -1)):
y_GT[start:end+1,1].fill(classesDict[gtList[i][1]])
# Check if we can write into the third column of the y_GT array:
elif ((len(np.unique(y_GT[start:end+1,2])) == 1) and
(np.unique(y_GT[start:end+1,2])[0] == -1)):
y_GT[start:end+1,2].fill(classesDict[gtList[i][1]])
# Check if we can write into the third column of the y_GT array:
elif ((len(np.unique(y_GT[start:end+1,3])) == 1) and
(np.unique(y_GT[start:end+1,3])[0] == -1)):
y_GT[start:end+1,3].fill(classesDict[gtList[i][1]])
# Check if we can write into the third column of the y_GT array:
elif ((len(np.unique(y_GT[start:end+1,4])) == 1) and
(np.unique(y_GT[start:end+1,4])[0] == -1)):
y_GT[start:end+1,4].fill(classesDict[gtList[i][1]])
else:
pdb.set_trace()
print("Problem occurred when filling ground truth array!")
break
if classesNotTrained:
for el in set(classesNotTrained):
print("The classifier wasn't trained with class '" +
el + "'. It will not be considered for testing.")
return y_GT