def computeKMeans(self):
"""Computes K means Clustering and returns the K centroids"""
#centroids, labels = scipy.cluster.vq.kmeans(self.patch_set,self.KMeans,self.Kmeans_iterations) #Compute Kmeans on the set of patches
centroids = cluster.k_means(self.patch_set,self.KMeans)
centroids = centroids[0]
for i in xrange(centroids.shape[0]): #Save these means as images to visualize (also enlarge for better visualiztion)
scipy.misc.imsave('CIFAR/Kmean/name'+str(i)+'.jpg', scipy.misc.imresize(centroids[i].reshape(3,PATCH_SIZE,PATCH_SIZE),(50,50)))
return None
def computeKMeans(self):
"""Computes K means Clustering and returns the K centroids"""
# centroids, labels = scipy.cluster.vq.kmeans(self.patch_set,self.KMeans,self.Kmeans_iterations)
centroids, labels, error = cluster.k_means(self.patch_set, self.KMeans)
for i in xrange(
centroids.shape[0]
): # Save these means as images to visualize (also enlarge for better visualiztion)
scipy.misc.imsave(
"Kaggle/Kmean/name" + str(i) + ".jpg",
scipy.misc.imresize(centroids[i].reshape(self.Patch_size, self.Patch_size, 3), (50, 50)),
)
return None
def whittenMeans(self):
"""Whittens the data and Computes K means Clustering and returns the K centroids"""
#pca_Obj = PCA(whiten=True) #Initialise the skLearn PCA decomposer for whittening the data
#transformed_patch_set = pca_Obj.fit_transform(self.patch_set.T)
#transformed_patch_set = pca_Obj.inverse_transform(transformed_patch_set)
transformed_patch_set = self.WhittenByPCA()
#centroids, labels = scipy.cluster.vq.kmeans(np.real(transformed_patch_set.T),self.KMeans,self.Kmeans_iterations) #K-means on this Whittened data
centroids = cluster.k_means(np.real(transformed_patch_set.T),self.KMeans)
centroids = centroids[0]
for i in xrange(centroids.shape[0]):
scipy.misc.imsave('CIFAR/Whitten/name'+str(i)+'.jpg', scipy.misc.imresize(centroids[i].reshape(3,PATCH_SIZE,PATCH_SIZE),(50,50)))
print 'Computed and Saved the means'
return None
def whittenMeans(self):
"""Whittens the data and Computes K means Clustering and returns the K centroids"""
pca_Obj = PCA(whiten=True) # Initialise the skLearn PCA decomposer for whittening the data
# transformed_patch_set = pca_Obj.fit_transform(self.patch_set.T)
transformed_patch_set = self.WhittenByPCA()
# centroids, labels = scipy.cluster.vq.kmeans(np.real(transformed_patch_set.T),self.KMeans,self.Kmeans_iterations) #K-means on this Whittened data
centroids, labels, error = cluster.k_means(np.real(transformed_patch_set.T), self.KMeans)
for i in xrange(centroids.shape[0]):
scipy.misc.imsave(
"Kaggle/Whitten/name" + str(i) + ".jpg",
scipy.misc.imresize(centroids[i].reshape(self.Patch_size, self.Patch_size, 3), (50, 50)),
)
print "Computed and Saved the means"
return None