def main():
datapath = '../TestData/'
predictionDir = '../Predictions/'
if not os.path.isdir(predictionDir):
os.mkdir(predictionDir)
Single = True
nbc = 128
binhalfwins = [2, 3, 4]
nbc = 128
lenthresh = 10
halfwins = [8, 17, 25]
alpha = 3
shiftstart = -4
shiftend = -1
fileList = os.listdir(datapath)
siz = 1
thresh = 0
samplelist = [files for files in fileList if files.startswith("Sample")]
for sample in samplelist:
if Single:
QuadDir = '{}{}/SpecificQuadDescriptors/'.format(datapath, sample)
else:
QuadDir = '{}{}/AllQuadDescriptors/'.format(datapath, sample)
print 'saving classifction number for', sample
smp = GestureSample(datapath, sample, training=False)
Numberofframe = smp.data['numFrames']
OutDir = '{}ClassifierOutputs/'.format(QuadDir)
allBinaryLabels = np.zeros((1, Numberofframe))
ss1 = 0
for binhalfwin in binhalfwins:
BinClfFile = BinClfFile = '{}binary_halfwin_{}.mat'.format(
OutDir, str(binhalfwin))
BinaryLabels = sio.loadmat(BinClfFile)['frame_label']
allBinaryLabels += BinaryLabels
ss1 += 1
allprobs = np.zeros((20, Numberofframe))
ss = 0
for halfwin in halfwins:
ss += 1
inFile = '{}Classfication_nbc_{}_halfwin{}.mat'.format(
OutDir, str(nbc), str(halfwin))
probabilties = 1 - np.transpose(sio.loadmat(inFile)['frame_probs'])
allprobs += probabilties
(r, c) = np.shape(probabilties)
M = np.zeros((r + 1, c))
M[:20, :] = allprobs / ss
# M[20,:] = BinaryLabels
BinaryLabels = mytools.myfilter(allBinaryLabels / ss1, siz, thresh)
# M[20,:] = binprob
M[20, :] = BinaryLabels
(p, D) = dpEM(M, alpha)
(labels, starts, ends) = getLabels(p)
FinalLabels = refineLabes(labels, starts, ends, M, lenthresh,
Numberofframe, shiftstart, shiftend)
tempfilename = '../Predictions/{}_prediction.csv'.format(sample)
mytools.exportPredictions(FinalLabels, tempfilename)
def main():
datapath = '../TestData/'
predictionDir = '../Predictions/'
if not os.path.isdir(predictionDir):
os.mkdir(predictionDir)
Single =True;
nbc =128
binhalfwins = [2,3,4];
nbc =128
lenthresh = 10
halfwins = [8,17,25];
alpha = 3
shiftstart = -4
shiftend = -1
fileList = os.listdir(datapath)
siz = 1;thresh = 0;
samplelist=[files for files in fileList if files.startswith("Sample")]
for sample in samplelist:
if Single:
QuadDir = '{}{}/SpecificQuadDescriptors/'.format(datapath,sample)
else:
QuadDir = '{}{}/AllQuadDescriptors/'.format(datapath,sample)
print 'saving classifction number for',sample
smp=GestureSample(datapath,sample,training = False);
Numberofframe = smp.data['numFrames'];
OutDir = '{}ClassifierOutputs/'.format(QuadDir)
allBinaryLabels = np.zeros((1,Numberofframe))
ss1=0
for binhalfwin in binhalfwins:
BinClfFile = BinClfFile = '{}binary_halfwin_{}.mat'.format(OutDir,str(binhalfwin))
BinaryLabels = sio.loadmat(BinClfFile)['frame_label']
allBinaryLabels += BinaryLabels
ss1+=1
allprobs = np.zeros((20,Numberofframe))
ss =0
for halfwin in halfwins:
ss+=1
inFile = '{}Classfication_nbc_{}_halfwin{}.mat'.format(OutDir,str(nbc),str(halfwin))
probabilties = 1-np.transpose(sio.loadmat(inFile)['frame_probs'])
allprobs += probabilties
(r,c) = np.shape(probabilties);
M = np.zeros((r+1,c));
M[:20,:] = allprobs/ss
# M[20,:] = BinaryLabels
BinaryLabels = mytools.myfilter(allBinaryLabels/ss1,siz,thresh)
# M[20,:] = binprob
M[20,:] = BinaryLabels;
(p,D) = dpEM(M,alpha)
(labels,starts,ends) = getLabels(p)
FinalLabels = refineLabes(labels,starts,ends,M,lenthresh,Numberofframe,shiftstart,shiftend)
tempfilename = '../Predictions/{}_prediction.csv'.format(sample)
mytools.exportPredictions(FinalLabels,tempfilename)
