def main():
filename = "data/SwedishLeaf_TRAIN"
data = open(filename, 'r').read()
data = data.split('\n')
print len(data)
features = np.zeros(shape=(len(data), 128))
for i, x in enumerate(data):
if (len(x) == 0):
continue
features[i] = (x.split(','))[1:]
distancesED = calc(features)
CFdistances = calcCF(features)
CID(500, 501, distancesED, CFdistances)
# print CFdistances
print "Now clustering"
from hcluster import hcluster
hcluster(features, distancesED, CFdistances)
def main():
filename = "data/SwedishLeaf_TRAIN"
data = open(filename, 'r').read()
data = data.split('\n')
print len(data)
features = np.zeros(shape=(len(data), 128))
for i, x in enumerate(data):
if (len(x) == 0):
continue
features[i] = (x.split(','))[1:]
distancesED = calc(features)
CFdistances = calcCF(features)
CID(500, 501, distancesED, CFdistances)
# print CFdistances
print "Now clustering"
from hcluster import hcluster
hcluster(features, distancesED, CFdistances)
def __init__(self, **kwargs):
if PARAM_OBS not in kwargs:
raise Exception(
"Cluster initialization requires initial observations")
elif PARAM_THRESHOLD not in kwargs:
raise Exception(
"Hierarchical cluster initialization requires a distance threshold."
)
obs_raw = kwargs[PARAM_OBS]
obs = numpy.array([arr[0] for arr in obs_raw])
self.adaptive = False
self.min_dist = float(kwargs[PARAM_THRESHOLD]) / 3
self.max_radius = float(kwargs[PARAM_THRESHOLD])
tree = hcluster(obs)
clusters = tree.extract_clusters(self.max_radius)
self.centroids = numpy.array([c.vec for c in clusters])
#trials, dist = self._cluster(obs_raw)
self.clusterNodes = [
ClusterNode(reduceDim(arr), 0.001) for arr in self.centroids
]
if PARAM_ADAPTIVE in kwargs:
self.adaptive = bool(kwargs[PARAM_ADAPTIVE])
#for idx in range(len(self.clusterNodes)):
# self.clusterNodes[trials[idx]].label = 'group {}'.format(idx)
return
def get_folder_clusters(path, imlist):
n = len(imlist)
#extract feature vector for each image
features = np.zeros((n, 3))
for i in range(n):
print 'working on ', imlist[i]
if 'average' in imlist[i]: continue
pathf = path + imlist[i]
im = np.array(Image.open(pathf))
if im.shape[-1] > 3:
im = to_rgb3b(im)
R = np.mean(im[:, :, 0].flatten())
G = np.mean(im[:, :, 1].flatten())
B = np.mean(im[:, :, 2].flatten())
features[i] = np.array([R, G, B])
tree = hcluster.hcluster(features)
print 'finished generating image clusters'
return tree
# -*- coding: utf-8 -*-
import sys
import hcluster
import find_files
import compute_hashes
import time
if __name__ == '__main__' and len(sys.argv) > 1:
print "searching for images..."
images = find_files.find_images(sys.argv[1])
images_length = len(images)
print images_length, "images found"
if images_length > 0:
hash_start_time = time.time()
print "computing hashes and vectors..."
hashes = compute_hashes.get_hashes(images)
hash_end_time = time.time()
print "all hashes computed in", hash_end_time - hash_start_time
print "clustering..."
hcluster.hcluster(hashes)
#!/usr/bin/env python3
import numpy
import pdb
import hcluster
list1 = [[292, 1203, 32, 17, 917], [830, 1079, 494], [46, 565, 262, 74, 21, 17, 539], [1050, 58, 52], [1079, 494], [16, 52], [16, 17], [29, 52], [405, 958, 292, 345, 97, 58, 17, 1061, 4, 494], [94, 446, 32, 1592, 194, 9392, 9, 687, 1540, 3029, 32, 52], [4687, 1741, 97, 32, 17, 238], [14, 15, 21, 52], [292, 1203, 32, 17, 917], [491, 48, 494], [806, 17], [1622, 16, 17], [16, 17], [46, 350, 97, 58, 494], [16, 17, 127, 2954], [379, 529, 15, 1516, 1531, 793, 17, 284], [29, 18708, 2687, 44, 1060, 17]]
length = len(list1)
jc_matrix = numpy.zeros((length,length) )
for i in range(length):
for j in range(i+1):
if i == j:
jc_matrix[i][j] = -1
else:
jc_matrix[i][j] = round(float(len(set(list1[i]) & set(list1[j])))/float(len(set(list1[i]) | set( list1[j]))),3) #round it 3 dec places
print jc_matrix
numpy.savetxt('jaccard.txt',jc_matrix, delimiter=" ", fmt="%s" )
tree = hcluster.hcluster(jc_matrix,list1)
# path = 'D:\\GitHub\\PCV-translation-to-Chinese\\data\\sunsets\\flickr-sunsets-small\\'
path = '../data/flickr-sunsets-small/'
imlist = [
os.path.join(path, f) for f in os.listdir(path) if f.endswith('.jpg')
]
# extract feature vector (8 bins per color channel)
features = zeros([len(imlist), 512])
for i, f in enumerate(imlist):
im = array(Image.open(f))
# multi-dimensional histogram
h, edges = histogramdd(im.reshape(-1, 3),
8,
normed=True,
range=[(0, 255), (0, 255), (0, 255)])
features[i] = h.flatten()
tree = hcluster.hcluster(features)
# visualize clusters with some (arbitrary) threshold
clusters = tree.extract_clusters(0.23 * tree.distance)
# plot images for clusters with more than 3 elements
# figure()
for c in clusters:
elements = c.get_cluster_elements()
nbr_elements = len(elements)
if nbr_elements > 3:
figure()
subplot(1, 2, 2)
for p in range(minimum(nbr_elements, 20)):
subplot(4, 5, p + 1)
im = array(Image.open(imlist[elements[p]]))
imshow(im)
# sort the edgelist
edgelist.sort()
featurematrix = []
temp = edgelist[0][0]
# print(edgelist[0])
# Generating the sparse matrix
fmline = []
for index, item in enumerate(edgelist):
newtemp = item[0]
if newtemp != temp:
temp = newtemp
featurematrix.append(fmline)
fmline = []
fmline.append(item[1])
else:
fmline.append(item[1])
# print(featurematrix)
# print(len(featurematrix))
"""
Building feature matrix from the edge list
extract feature vector for each node
"""
# tree = hclusttest.hcluster(featurematrix)
tree = hcluster.hcluster(featurematrix)
def main(jobName, inputfile):
f = open(inputfile)
natoms, SDdata, DistAvg, labels = readInput(f)
f.close()
tree = hcluster.hcluster(natoms, labels, SDdata, DistAvg)
writeOutput(natoms, tree)
# PCA on all images.
imlist = imtools.get_imlist('/Users/thakis/Downloads/data/a_thumbs')
imcount = len(imlist)
immatrix = array([array(Image.open(im)).flatten() for im in imlist], 'f')
V, S, immean = pca.pca(immatrix)
# Visualize only selected images.
imlist = imtools.get_imlist('/Users/thakis/Downloads/data/a_selected_thumbs')
imcount = len(imlist)
immatrix = array([array(Image.open(im)).flatten() for im in imlist], 'f')
# Project on 40 first PCs.
projected = array([dot(V[:40], immatrix[i] - immean) for i in range(imcount)])
# hierarchical clustering.
tree = hcluster.hcluster(projected)
hcluster.draw_dendrogram(tree, imlist, filename='out_font.png')
# k means.
K = 4
projected = whiten(projected)
centroids, variance = kmeans(projected, K)
code, distance = vq(projected, centroids)
# Plot clusters.
for k in range(K):
ind = where(code == k)[0]
figure()
gray()
for i in range(minimum(len(ind), 40)):
subplot(4, 10, i + 1)
# -*- coding: utf-8 -*-
"""
Created on Fri Sep 02 16:39:06 2016
@author: user
"""
import hcluster
from numpy import *
from PIL import Image
import os
#create the dataset
class1 = 1.5 * randn(100, 2)
class2 = randn(100, 2) + array([5, 5])
features = vstack((class1, class2))
tree = hcluster.hcluster(features)
clusters = tree.extract_clusters(5)
print len(clusters)
for c in clusters:
print c.get_cluster_elements()
#use the images features named sunset
# create a list of images
path = '../pcv_data/data/sunsets/flickr-sunsets-small/'
imlist = [
os.path.join(path, f) for f in os.listdir(path) if f.endswith('.jpg')
]
# extract feature vector (8 bins per color channel)
features = zeros([len(imlist), 512])
for i, f in enumerate(imlist):
def draw_tree(observations, filename):
node = hcluster([o[0] for o in observations])
draw_dendrogram(node, [o[1] for o in observations], filename)
return filename
for i in range(len(dataset)):
feaArr = dataset[i][:,3:9]
timestamp = dataset[i][:,0]
filtfea = filtSensordata(feaArr)
timestamp = map(lambda x:float(x),timestamp)
timestampArr = np.array([timestamp]).T
sensordata = np.append(timestampArr,filtfea,1)
activities = detectActivity(sensordata)
print "the %dth file contains:%d activities"%(i,len(activities))
activitiesofAll += activities
print "activity number:%d" % len(activitiesofAll)
activitiesofMini = divideMiniActivity(activitiesofAll)
activitiesfea = feaExtraction(activitiesofMini)
n_clusters = 3
#x=array([[[1,1,1],[2,2,2],[3,3,3]],[[1,1,1],[2,2,2],[3,3,3]],[[4,4,4],[5,5,5],[6,6,6]],[[4,4,4],[5,5,5],[6,6,6]]])
k,l = hcluster(activitiesfea,4)
print l
#print len(timestampArr),len(feaArr)
'''
rotationArr = getRotationArr(timestamp,filtfea)
plt.figure(1)
ax1 = plt.subplot(311)
plt.ylabel(u'rotationx')
plt.xlabel(u'sample')
plt.plot(range(len(rotationArr)),rotationArr[:,0],color = 'red',linewidth = 1.0)
plt.xlim(0,len(rotationArr))
plt.ylim(-200,200)
ax2 = plt.subplot(312)
plt.ylabel(u'rotationy')
#クラスタのラベルを使って画像を生成する
codeim = code.reshape(steps,steps)
codeim = imresize(codeim,im.shape[:2],interp='nearest')
plt.figure()
plt.imshow(codeim)
#この後、projectedとimlistの情報を利用するので、一時的に違う名前で保存しておく
projected2=projected
imlist2=imlist
#6.2 階層クラスタリング
class1 = 1.5 * np.random.randn(100,2)
class2 = np.random.randn(100,2) + np.array([5,5])
features = np.vstack((class1,class2))
tree = hcluster.hcluster(features)
clusters = tree.extract_clusters(5)
print 'number of clusters', len(clusters)
for c in clusters:
print c.get_cluster_elements()
#6.2.1 画像のクラスタリング
#画像のリストを作成する
path = 'flickr-sunsets/'
imlist = [os.path.join(path,f) for f in os.listdir(path) if f.endswith('.jpg')]
#特徴量ベクトルを抽出する（色チャンネルに８つのビン）
features = np.zeros([len(imlist), 512])
from PIL import Image
from pylab import *
import imtools
import pickle
from scipy.cluster.vq import *
# 画像のリストを得る
imlist = imtools.get_imlist('selected_fontimages/')
imnbr = len(imlist)
# モデルのファイルを読み込む
with open('font_pca_modes.pkl','rb') as f:
immean = pickle.load(f)
V = pickle.load(f)
# 平板化した画像を格納する行列を作る
immatrix = array([array(Image.open(im)).flatten()
for im in imlist],'f')
# 第40主成分までを射影する
immean = immean.flatten()
projected = array([dot(V[:40],immatrix[i]-immean)
for i in range(imnbr)])
import hcluster
tree = hcluster.hcluster(projected)
hcluster.draw_dendrogram(tree,imlist,filename='fonts.jpg')
from pylab import * class1 = 1.5 * randn(100,2) class2 = randn(100,2) + array([5,5]) features = vstack((class1,class2)) import hcluster tree = hcluster.hcluster(features) clusters = tree.extract_clusters(5) print 'number of clusters', len(clusters) for c in clusters: print c.get_cluster_elements()
def jcsimilarity(ne1,ne2,conveclist,docs):
length = len(conveclist)
temp_list = []
jc_matrix=[]
for i in range(length):
for j in range(i+1):
if i == j:
temp_list.extend([0])
else:
temp_list.extend([round(float(len(set(conveclist[i]) & set(conveclist[j])))/float(len(set(conveclist[i]) | set( conveclist[j]))),3)])
jc_matrix.append(temp_list)
temp_list=[]
numpy.savetxt('jaccard2.txt',jc_matrix, delimiter=" ", fmt="%s" )
jcc_matrix = [[0],[10,0],[2,4,0],[3,6,5,0],[1,2,3,9,0],[6,7,8,20,2,0]]
# print ('Enter hcluster')
relations = hcluster.hcluster(jc_matrix, conveclist)
# print ('Exit hcluster')
relations.sort()
if not relations:
return
relation = []
fp = open('wordlist.pkl')
wordlist = pickle.load(fp)
inv_wordlist = {v : k for k, v in wordlist.items()}
fp.close()
fp = open('NamedEntity.pkl')
ne = pickle.load(fp)
inv_ne = {v : k for k , v in ne.items()}
fp.close()
prev = []
# count = 0
if (len(relations)) > 1:
while len(relations) > 0 :
rel = relations.pop()
if rel != prev:
if prev != []:
try:
with open('Relation.csv', 'a') as myfile:
if len(prev) > 0:
writer = csv.writer(myfile, dialect='excel')
row = [inv_ne[ne1],inv_ne[ne2],r,docs] # change inv_wordlist with relation
writer.writerow(row)
except IOError as ioe:
print('Error: ' + str(ioe))
#relation = []
# count = 0
r = []
prev = rel[:]
while len(rel) > 0:
r.append(inv_wordlist[rel.pop()])
# relation.append(r)
# count += 1
# else:
# r = []
# prev = rel[:]
# while len(rel) > 0:
# r.append(inv_wordlist[rel.pop()])
# relation.append(r)
# count += 1
try:
with open('Relation.csv', 'a') as myfile:
if len(prev) > 0 :
writer = csv.writer(myfile, dialect='excel')
row = [inv_ne[ne1],inv_ne[ne2], r,docs]
writer.writerow(row)
except IOError as ioe:
print('Error: ' + str(ioe))