def build_codebook(self, k):
print('Building a GMM of {} components as a codebook'.format(k))
return gaussian_mixture.GaussianMixture(n_components=k,
verbose=False,
covariance_type='diag',
tol=1e-3,
reg_covar=1e-6,
max_iter=100)
def em(data):
# Initialize the clusterer to make 3 clusters
# seed of 10 for reproducibility.
clusterer = gaussian_mixture.GaussianMixture(n_components=3,
covariance_type="spherical")
train, test = seperateData(data)
clusterer.fit(train) #train
cluster_labels = clusterer.predict(test) #test
cluster_labels_list = cluster_labels.tolist()
#print(cluster_labels_list)
return cluster_labels_list
def cluster_image(mic, n_comps):
hsv = matplotlib.colors.rgb_to_hsv(mic.data.cpu().permute(1, 2, 0).numpy())
rgb = mic.data.cpu().permute(1, 2, 0).numpy()
comb_data = np.concatenate([rgb, hsv], -1)
newdata = comb_data.reshape(mic.data.shape[1] * mic.data.shape[2], 6)
model = gmm.GaussianMixture(n_components=n_comps, covariance_type="full")
model = model.fit(newdata)
cluster = model.predict(newdata)
cluster = cluster.reshape(mic.data.shape[1], mic.data.shape[2])
return cluster
def train_model(user_name):
warnings.filterwarnings(action="ignore", category=DeprecationWarning)
# DATASET_PATH = "DATASET"
#fhand=open("user_name.txt",mode='r+')
features = np.asarray(())
dest = "user_models/"
a = user_name
for dirpath, dirnames, filenames in os.walk(dest):
# print(filenames)
temp_name = a + ".gmm"
if temp_name in filenames:
# print("already exist")
sys.exit()
count = 1
DATASET_PATH = "DATASET/" + a
for dirpath, dirnames, filenames in os.walk(DATASET_PATH):
for f in filenames:
file_path = os.path.join(dirpath, f)
#read audio file
sr, audio = read(file_path)
vector = extract_features(audio, sr)
if features.size == 0:
features = vector
else:
features = np.vstack((features, vector))
if count == 5:
gmm = gaussian_mixture.GaussianMixture(n_components=16,
max_iter=200,
covariance_type='diag',
n_init=3)
gmm.fit(features)
# dumping the trained gaussian model
picklefile = a + ".gmm"
# print("model name ",picklefile)
file = open(dest + picklefile, mode="wb")
pickle.dump(gmm, file)
# print ('+ modeling completed for speaker:',picklefile," with data point = ",features.shape)
features = np.asarray(())
count = 0
count = count + 1
# print("gathering ROOT SIFT descriptors...")
# descriptors=fun.compute_save_reduce_vector(paths,id,pc_comp=pc_comp,reduced=True).T
descriptors = np.atleast_2d(
np.asarray(
fun.file_counter(paths,
".npy",
"reduced_data",
remove=False,
loader=True)))
# print("descriptors gathered")
print("training GMM %d..." % (gmm_comp))
#GMM MODEL
GMM = gauss.GaussianMixture(n_components=gmm_comp,
covariance_type=covariance_type,
max_iter=100000,
n_init=1,
init_params="kmeans")
GMM.fit(descriptors)
# print(np.sum(GMM.predict_proba(descriptors[0:20]),axis=1))
print("trained GMM %d..." % (gmm_comp))
print("saving the GMM model")
means = GMM.means_
covs = GMM.covariances_
weights = GMM.weights_
gmm_means_file = "./GMM/means" + str(gmm_comp) + ".gmm.npy"
gmm_covariance_file = "./GMM/covs" + str(gmm_comp) + ".gmm.npy"
gmm_weight_file = "./GMM/weights" + str(gmm_comp) + ".gmm.npy"
np.save(gmm_means_file, means)
def GMM(x, i):
gmm = gaussian_mixture.GaussianMixture(n_components=i).fit(x)
labels = gmm.predict(x)
plt.scatter(x[:, 0], x[:, 1], c=labels, cmap='viridis')
plt.show()
max_iter=1000,
n_init=10,
random_state=0,
verbose=False,
tol=1e-4) # K-Means Algorithm
kmeans.fit(X)
y_kmeans = kmeans.predict(X) # Prediction using K-Means
#print(kmeans.labels_)
a = silhouette_score(X, kmeans.labels_)
sil.append(a)
wcss.append(
kmeans.inertia_
) #appending the sum of squared distances to the cluster center
gmm = GM.GaussianMixture(n_components=j,
init_params='kmeans',
max_iter=1000,
covariance_type='full',
tol=1e-04,
random_state=0) # Gaussian Mixture Modelling
gmm.fit(X)
bics.append(gmm.bic(X)) # updating the BIC list
index_gmm = np.argmin(bics) # index of minimum BIC penalty
index_sil = np.argmax(sil)
#Plotting cost vs number of clusters
plt.figure("K-Means Clustering Analysis using Elbow Method")
plt.plot(K, wcss, 'go--')
plt.title('The Elbow Method')
plt.xlabel('Number of clusters')
plt.ylabel('inertia')
plt.show()
#plt.show()
import matplotlib.pyplot as plt
from sklearn.datasets import samples_generator
from sklearn import metrics, cluster
from sklearn.mixture import gaussian_mixture
# x,y = samples_generator.make_blobs(n_samples=200,n_features=3,cluster_std=0.6,random_state=0)
x, y = samples_generator.make_circles(n_samples=200,
noise=.05,
random_state=0,
factor=0.4)
# x,y = samples_generator.make_moons(n_samples=200,noise=.05,random_state=0)
# print(x.shape,y.shape)
# clu = cluster.KMeans(2)
# clu = cluster.MeanShift()
# clu = cluster.DBSCAN(eps=0.98,min_samples=4)
# clu = cluster.SpectralClustering(2,affinity="nearest_neighbors")
# clu = cluster.AffinityPropagation()
clu = gaussian_mixture.GaussianMixture(n_components=2)
labels = clu.fit_predict(x)
print(metrics.silhouette_score(x, labels))
print(metrics.calinski_harabasz_score(x, labels))
print(metrics.davies_bouldin_score(x, labels))
plt.scatter(x[:, 0], x[:, 1], c=labels)
plt.show()
data1 = np.random.multivariate_normal(miu1, cov1, size=n) data2 = np.random.multivariate_normal(miu2, cov2, size=n) data3 = np.random.multivariate_normal(miu3, cov3, size=n) data = np.concatenate((data1, data2, data3)) np.random.shuffle(data) print(data.shape) plt.xlim(-40, 50) plt.ylim(-40, 50) start_time = time.time() gmm = gaussian_mixture.GaussianMixture(n_components=3) gmm.fit(data) data_labels = gmm.predict(data) test_labels = gmm.predict(test_data) end_time = time.time() print((end_time - start_time)) f = plt.figure(1) plt.scatter(data[:, 0], data[:, 1], c=data_labels, s=20, cmap='viridis'); # plt.plot(data[:, 0], data[:, 1], marker="o", linestyle="") # plt.show()
def showSilhouette(data, cluster_labels, attributes):
X = []
for instance in data:
X.append([instance[attributes[0]], instance[attributes[1]]])
X = np.array(X)
#print(X)
# Create a subplot with 1 row and 2 columns
fig, (ax1, ax2) = plt.subplots(1, 2)
fig.set_size_inches(13, 5.75)
# The 1st subplot is the silhouette plot
# The silhouette coefficient can range from -1, 1 but in this example all
# lie within [-0.1, 1]
ax1.set_xlim([-0.1, 1])
n_clusters = 3
# The (n_clusters+1)*10 is for inserting blank space between silhouette
# plots of individual clusters, to demarcate them clearly.
ax1.set_ylim([0, len(X) + (n_clusters + 1) * 10])
# The silhouette_score gives the average value for all the samples.
# This gives a perspective into the density and separation of the formed
# clusters
clusterer_s = gaussian_mixture.GaussianMixture(n_components=3,
covariance_type="spherical")
clusterer_s.fit(X)
cluster_labels_s = clusterer_s.predict(X)
colors_scatter = []
colorT = ""
for value in cluster_labels_s:
if (value == 0):
colorT = "red"
if (value == 1):
colorT = "blue"
if (value == 2):
colorT = "green"
colors_scatter.append(colorT)
#print(cluster_labels_s)
silhouette_avg = silhouette_score(X, cluster_labels_s)
print("For 3 clusters the average silhouette_score is:",
round(silhouette_avg, 5))
# Compute the silhouette scores for each sample
sample_silhouette_values = silhouette_samples(X, cluster_labels_s)
y_lower = 10
for i in range(n_clusters):
# Aggregate the silhouette scores for samples belonging to
# cluster i, and sort them
ith_cluster_silhouette_values = \
sample_silhouette_values[cluster_labels_s == i]
ith_cluster_silhouette_values.sort()
#print(ith_cluster_silhouette_values)
size_cluster_i = ith_cluster_silhouette_values.shape[0]
y_upper = y_lower + size_cluster_i
if (i == 0):
color = "red"
if (i == 1):
color = "blue"
if (i == 2):
color = "green"
#print(y_lower)
#print(y_upper)
#print(np.arange(y_lower, y_upper))
ax1.fill_betweenx(np.arange(y_lower, y_upper),
0,
ith_cluster_silhouette_values,
facecolor=color,
edgecolor=color,
alpha=1)
# Label the silhouette plots with their cluster numbers at the middle
ax1.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
# Compute the new y_lower for next plot
y_lower = y_upper + 10 # 10 for the 0 samples
ax1.set_title("The silhouette plot for the various clusters.")
ax1.set_xlabel("The silhouette coefficient values")
ax1.set_ylabel("Cluster label")
# The vertical line for average silhouette score of all the values
ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
ax1.set_yticks([]) # Clear the yaxis labels / ticks
ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
# 2nd Plot showing the actual clusters formed
colors = ["red", "blue", "green"]
#print(type(X))
#print(colors)
ax2.scatter(X[:, 0],
X[:, 1],
marker='.',
s=50,
lw=0,
alpha=1,
c=colors_scatter,
edgecolor='k')
# Labeling the clusters
centers = clusterer_s.means_
# Draw white circles at cluster centers
ax2.scatter(centers[:, 0],
centers[:, 1],
marker='o',
c="white",
alpha=1,
s=200,
edgecolor='k')
for i, c in enumerate(centers):
ax2.scatter(c[0],
c[1],
marker='$%d$' % i,
alpha=1,
s=70,
edgecolor='k')
#print(clusterer_s.means_)
ax2.set_title("The visualization of the clustered data.")
ax2.set_xlabel("Feature space for the 1st feature")
ax2.set_ylabel("Feature space for the 2nd feature")
plt.suptitle(("Silhouette analysis for EM clustering on seed data "
"with n_clusters = %d" % n_clusters),
fontsize=14,
fontweight='bold')
plt.show()