def test_non_contiguous_fmatrix():
from milksets.wine import load
features,_ = load()
features = features[:,::2]
assigns, centroids = milk.unsupervised.kmeans(features, 3, R=2, max_iters=10)
assert np.all(assign_centroids(features, centroids) == assigns)
features = features.astype(np.int32)
assigns, centroids = milk.unsupervised.kmeans(features, 3, R=2, max_iters=10)
assert np.all(assign_centroids(features, centroids) == assigns)
def test_non_contiguous_fmatrix():
from milksets.wine import load
features,_ = load()
features = features[:,::2]
assigns, centroids = milk.unsupervised.kmeans(features, 3, R=2, max_iters=10)
assert np.all(assign_centroids(features, centroids) == assigns)
features = features.astype(np.int32)
assigns, centroids = milk.unsupervised.kmeans(features, 3, R=2, max_iters=10)
assert np.all(assign_centroids(features, centroids) == assigns)
def apply(self, features):
from milk.unsupervised.kmeans import assign_centroids
f0,f1 = features
features = assign_centroids(f0, self.centroids, histogram=True, normalise=self.normalise)
if f1 is not None and len(f1):
features = np.concatenate((features, f1))
return self.base.apply(features)
def train(self, features, labels, **kwargs):
from milk.unsupervised.kmeans import assign_centroids
tfeatures = np.array([ assign_centroids(f, self.codebook, histogram=True, normalise=self.normalise)
for f,_ in features])
tfeatures = np.hstack((tfeatures, np.array([f for _,f in features])))
base_model = self.base.train(tfeatures, labels, **kwargs)
return codebook_model(self.codebook, base_model, self.normalise)
def apply(self, features):
from milk.unsupervised.kmeans import assign_centroids
f0,f1 = features
features = assign_centroids(f0, self.centroids, histogram=True, normalise=self.normalise)
if f1 is not None and len(f1):
features = np.concatenate((features, f1))
return self.base.apply(features)
def train(self, features, labels, **kwargs):
from milk.unsupervised.kmeans import assign_centroids
tfeatures = np.array([ assign_centroids(f, self.codebook, histogram=True, normalise=self.normalise)
for f,_ in features])
tfeatures = np.hstack((tfeatures, np.array([f for _,f in features])))
base_model = self.base.train(tfeatures, labels, **kwargs)
return codebook_model(self.codebook, base_model, self.normalise)
def project(features, centroids):
from milk.unsupervised.kmeans import assign_centroids
return np.array([
np.concatenate([
assign_centroids(s, centroids, histogram=True, normalise=True),
other
])
for s,other in features])
def test_assign_cids():
from milksets.wine import load
features, _ = load()
assigns, centroids = milk.unsupervised.kmeans(features,
3,
R=2,
max_iters=10)
assert np.all(assign_centroids(features, centroids) == assigns)
def aic(features, centroids):
from milk.unsupervised.gaussianmixture import AIC
from milk.unsupervised.kmeans import assign_centroids
assignments = []
feats = []
for fs,_ in features:
assignments.extend(assign_centroids(fs, centroids))
feats.append(fs)
return AIC(np.concatenate(feats), assignments, centroids)
def train(self, features, labels, **kwargs):
allfeatures = np.concatenate(features)
assignments, centroids = select_best_kmeans(allfeatures, self.ks, 1,
"AIC")
histograms = [
assign_centroids(f, centroids, histogram=True, normalise=1)
for f in features
]
base_model = self.base.train(histograms, labels, **kwargs)
return precluster_model(centroids, base_model)
def train(self, features, labels, **kwargs):
allfeatures = np.vstack(features)
assignments, centroids = select_best_kmeans(allfeatures,
self.ks,
repeats=1,
method="AIC",
R=self.R)
histograms = [
assign_centroids(f,
centroids,
histogram=True,
normalise=self.normalise) for f in features
]
base_model = self.base.train(histograms, labels, **kwargs)
return precluster_model(centroids, base_model)
def testing(self, qImages):
# set count for each brand to 0
t01=0
t11=0
t21=0
t31=0
t41=0
for qImage in qImages:
# read input image
print qImage
qImg = imread(qImage)
keypoints, descriptors = self.extractFeatures(qImg)
descriptors = np.array(descriptors)
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 10, 1.0)
#scikit learn clusters
t00=time()
self.assignClusters(descriptors)
t01 = t01+ time()-t00
FLANN_INDEX_KDTREE = 0
index_params = dict(algorithm = FLANN_INDEX_KDTREE, trees = 5)
search_params = dict(checks=50)
# Flann matcher
t1=time()
flann = cv2.FlannBasedMatcher(index_params,search_params)
flann.knnMatch(descriptors,self.opencv_centers,k=1)
t11 = t11+ time()-t1
# brute Force Matcher
t2=time()
bf = cv2.BFMatcher(cv2.NORM_L2, crossCheck=False)
bf.knnMatch(descriptors,self.opencv_centers,k=1)
t21 = t21+ time()-t2
#scipy Kmeans
t3= time()
kmeans2(descriptors, self.scipy_centers, iter=10)
t31 = t31+ time()-t3
# milk cluster
t4=time()
assign_centroids(descriptors, self.milk_centers)
t41 = t41+ time()-t4
print "time for scikit learn to assign cluster",t01
print "time for flann matcher to assign clusters",t11
print "time for bfmatcher to assign clusters",t21
print "time for scipy kmeans to assign clusters ",t31
print "time for milk with parallel processing copyrights MIT to assign clusters ",t41
def train(self, features, labels, **kwargs):
allfeatures = np.concatenate(features)
assignments, centroids = select_best_kmeans(allfeatures, self.ks, 1, "AIC")
histograms = [assign_centroids(f, centroids, histogram=True, normalise=1) for f in features]
base_model = self.base.train(histograms, labels, **kwargs)
return precluster_model(centroids, base_model)
def train(self, features, labels, **kwargs):
allfeatures = np.vstack(features)
assignments, centroids = select_best_kmeans(allfeatures, self.ks, repeats=1, method="AIC", R=self.R)
histograms = [assign_centroids(f, centroids, histogram=True, normalise=self.normalise) for f in features]
base_model = self.base.train(histograms, labels, **kwargs)
return precluster_model(centroids, base_model)
def apply(self, features):
histogram = assign_centroids(features, self.centroids, histogram=True, normalise=self.normalise)
return self.base.apply(histogram)
def apply(self, features):
histogram = assign_centroids(features, self.centroids, histogram=True, normalise=1)
return self.base.apply(histogram)
def test_assign_cids():
from milksets.wine import load
features,_ = load()
assigns, centroids = milk.unsupervised.kmeans(features, 3, R=2, max_iters=10)
assert np.all(assign_centroids(features, centroids) == assigns)
