def createDataset(svmName='',roiName='',labFiles=[],niiFiles=[],lab1='',lab2=''):
"""
Workhorse...
"""
if os.path.exists(svmName):
print('Overwriting {0}.'.format(svmName))
os.remove(svmName)
for labF, niiF in zip(labFiles, niiFiles):
print(niiF,labF)
## Get the needed data
vtc = nf.NiftiImage(niiF)
roi = nf.NiftiImage(roiName)
vols, labels = pre.readLabList(labF)
## Preprocess the data
reducedRoi = pre.roiReduce(roi,vtc)
maskedVtc = pre.maskVtc(vtc,reducedRoi)
reference = pre.createRefVtc(maskedVtc)
## Filter labels and vols by trainLab1, trainLab2
## then change recode the labels as 1 and 2
l1mask = labels == lab1
l2mask = labels == lab2
l1l2mask = l1mask != l2mask
labels[l1mask] = 1
labels[l2mask] = 2
vols = vols[l1l2mask]
labels = labels[l1l2mask]
pre.writeSVM(maskedVtc,reference,labels,vols,svmName)
else:
print('Z-normalizing the data.')
pre.zSparse(svmName)
#pre.writeSVM(maskedVtc,reference,labels,vols,svmName)
print('Creating testing data:')
for labF, niiF in zip(testLabsFiles, testNiiFiles):
print(niiF,labF)
## Get the needed data
vtc = nf.NiftiImage(niiF)
roi = nf.NiftiImage(roiVmr)
vols, labels = pre.readLabList(labF)
## Preprocess the data
reducedRoi = pre.roiReduce(roi,vtc)
maskedVtc = pre.maskVtc(vtc,reducedRoi)
reference = pre.createRefVtc(maskedVtc)
## Filter labels and vols by trainLab1, trainLab2
## then change recode the labels as 1 and 2
l1mask = labels == testLab1
l2mask = labels == testLab2
l1l2mask = l1mask != l2mask
labels[l1mask] = 1
labels[l2mask] = 2
vols = vols[l1l2mask]
labels = labels[l1l2mask]
svmName = 'test_'+ testLab1 + 'x' + testLab2 + '_' + svmBaseName
pre.writeSVM(maskedVtc,reference,labels,vols,svmName)
# Add a timestamp to the SVM file name, preventing accidental
# overwriting or modification.
print('Creating training data:')
for labF, niiF in zip(trainLabsFiles, trainNiiFiles):
print(niiF,labF)
## Get the needed datword_a
vtc = nf.NiftiImage(niiF)
roi = nf.NiftiImage(roiVmr)
vols, labels = pre.readLabList(labF)
## Preprocess the data
reducedRoi = pre.roiReduce(roi,vtc)
maskedVtc = pre.maskVtc(vtc,reducedRoi)
reference = pre.createRefVtc(maskedVtc)
## Filter labels and vols by trainLab1, trainLab2
## then change recode the labels as 1 and 2
l1mask = labels == trainLab1
l2mask = labels == trainLab2
l1l2mask = l1mask != l2mask
labels[l1mask] = 1
labels[l2mask] = 2
vols = vols[l1l2mask]
labels = labels[l1l2mask]
svmName = 'train_scene_'+ trainLab1 + 'x' + trainLab2 + '_' + svmBaseName
pre.writeSVM(maskedVtc,reference,labels,vols,svmName)