def train(self, C=None, Gamma=None, ilog=None, callback=None):
# Create the SVM
self.svm = SVM(svm_type=TYPE_C_SVC)
# Add training data
for sub_id, tiles in self.training_data.iteritems():
for tile in tiles:
self.svm.addTraining(sub_id, tile)
# Train the SVM
if C != None and Gamma != None:
self.svm.train(C_range=C,
G_range=Gamma,
verbose=True,
callback=callback)
else:
#Automatic
self.svm.train(verbose=True, callback=callback)
def train(self, C = None, Gamma = None, ilog=None, callback=None):
# Create the SVM
self.svm = SVM(type=TYPE_C_SVC)
# Add training data
for id, tiles in self.training_data.iteritems():
for tile in tiles:
self.svm.addTraining(id,tile)
# Train the SVM
if C != None and Gamma != None:
self.svm.train(C_range = C, G_range = Gamma, verbose=True,callback=callback)
else:
#Automatic
self.svm.train(verbose=True,callback=callback)
def testSVM():
svm = SVM()
for i in iris_training:
svm.addTraining(labels[i], iris_data[i, :])
svm.train(verbose=1)
success = 0.0
total = 0.0
for i in iris_testing:
c = svm.predict(iris_data[i, :])
#print c, labels[i]
if c == labels[i]:
success += 1
total += 1
print "SVM Rate:", success / total
def testSVM():
svm = SVM()
for i in iris_training:
svm.addTraining(labels[i],iris_data[i,:])
svm.train(verbose = 1)
success = 0.0
total = 0.0
for i in iris_testing:
c = svm.predict(iris_data[i,:])
#print c, labels[i]
if c == labels[i]:
success += 1
total += 1
print "SVM Rate:",success/total
table = pv.Table()
values = {0: [], 1: []}
correct = 0
total = 0
for each in testing:
label = clsfy.predict(each[1], ilog=ilog)
total += 1
if label == each[0]:
correct += 1
rate = float(correct) / total
if ilog: ilog.table(table)
return rate
if __name__ == "__main__":
from pyvision.vector.SVM import SVM
svm = SVM(kernel='LINEAR', random_seed=30)
ilog = pv.ImageLog()
print "SVM rate:", genderClassifier(svm, ilog=None)
svm = SVM(kernel='RBF', random_seed=30)
ilog = pv.ImageLog()
print "SVM rate:", genderClassifier(svm, ilog=None)
ilog.show()
class SVMFaceRec:
def __init__(self):
self.tile_size = (128,160)
self.leye = pv.Point(26.0,40.0)
self.reye = pv.Point(102.0,40.0)
self.norm_sigma = 8.0
self.svm = None
self.n_faces = 0
self.n_labels = 0
self.training_data = {}
def preprocess(self,im,leye,reye,ilog=None):
im = pv.Image(im.asPIL())
affine = pv.AffineFromPoints(leye,reye,self.leye,self.reye,self.tile_size)
tile = affine.transformImage(im)
mat = tile.asMatrix2D()
# High pass filter the image
mat = mat - ndimage.gaussian_filter(mat,self.norm_sigma)
# Value normalize the image.
mat = mat - mat.mean()
mat = mat / mat.std()
tile = pv.Image(mat)
return tile
def addTraining(self,im,leye,reye,id,ilog=None):
self.svm = None
tile = self.preprocess(im,leye,reye,ilog)
if not self.training_data.has_key(id):
self.training_data[id] = []
self.training_data[id].append(tile)
self.n_labels = len(self.training_data)
self.n_faces += 1
def train(self, C = None, Gamma = None, ilog=None, callback=None):
# Create the SVM
self.svm = SVM(type=TYPE_C_SVC)
# Add training data
for id, tiles in self.training_data.iteritems():
for tile in tiles:
self.svm.addTraining(id,tile)
# Train the SVM
if C != None and Gamma != None:
self.svm.train(C_range = C, G_range = Gamma, verbose=True,callback=callback)
else:
#Automatic
self.svm.train(verbose=True,callback=callback)
def predict(self,im,leye,reye):
assert self.svm != None
tile = self.preprocess(im,leye,reye)
return self.svm.predict(tile)
def score(self,im,leye,reye):
tile = self.preprocess(im,leye,reye)
return self.svm.predictSVMValues(tile)
def reset(self):
self.svm = None
self.n_faces = 0
self.n_labels = 0
self.training_data = {}
def isTrained(self):
return self.svm != None
def runSVM(self):
svm = SVM()
print "I am in the SVM module now"
class SVMFaceRec:
def __init__(self):
self.tile_size = (128, 160)
self.leye = pv.Point(26.0, 40.0)
self.reye = pv.Point(102.0, 40.0)
self.norm_sigma = 8.0
self.svm = None
self.n_faces = 0
self.n_labels = 0
self.training_data = {}
def preprocess(self, im, leye, reye, ilog=None):
im = pv.Image(im.asPIL())
affine = pv.AffineFromPoints(leye, reye, self.leye, self.reye,
self.tile_size)
tile = affine.transformImage(im)
mat = tile.asMatrix2D()
# High pass filter the image
mat = mat - ndimage.gaussian_filter(mat, self.norm_sigma)
# Value normalize the image.
mat = mat - mat.mean()
mat = mat / mat.std()
tile = pv.Image(mat)
return tile
def addTraining(self, im, leye, reye, sub_id, ilog=None):
self.svm = None
tile = self.preprocess(im, leye, reye, ilog)
if not self.training_data.has_key(sub_id):
self.training_data[sub_id] = []
self.training_data[sub_id].append(tile)
self.n_labels = len(self.training_data)
self.n_faces += 1
def train(self, C=None, Gamma=None, ilog=None, callback=None):
# Create the SVM
self.svm = SVM(svm_type=TYPE_C_SVC)
# Add training data
for sub_id, tiles in self.training_data.iteritems():
for tile in tiles:
self.svm.addTraining(sub_id, tile)
# Train the SVM
if C != None and Gamma != None:
self.svm.train(C_range=C,
G_range=Gamma,
verbose=True,
callback=callback)
else:
#Automatic
self.svm.train(verbose=True, callback=callback)
def predict(self, im, leye, reye):
assert self.svm != None
tile = self.preprocess(im, leye, reye)
return self.svm.predict(tile)
def score(self, im, leye, reye):
tile = self.preprocess(im, leye, reye)
return self.svm.predictSVMValues(tile)
def reset(self):
self.svm = None
self.n_faces = 0
self.n_labels = 0
self.training_data = {}
def isTrained(self):
return self.svm != None
