/
Umap-Test1.py
208 lines (166 loc) · 6.13 KB
/
Umap-Test1.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
# -*- coding: utf-8 -*-
"""
UMAP plot for input images clustered according to similarity.
Created on Sun Dec 12 11:28:45 2021
@author: jbeale
https://umap-learn.readthedocs.io/en/latest/parameters.html
"""
import numpy as np
import seaborn as sns
import umap
import umap.plot
from math import sqrt
import os # loop over images in directory
import skimage.io # to read & save images
from skimage.util import img_as_float
import pandas as pd
# ---------------------------------------------------------------
# cartesian distance between two points (a,b are 2-elem vectors)
def jdist(a, b):
dx = a[0] - b[0]
dy = a[1] - b[1]
dist = sqrt(dx*dx + dy*dy)
return(dist)
# -------------------------------------------------------
# find index and distance of the nearest elment in ar[] to point p
def jmin(p, ar):
mdist = jdist(p, ar[0])
nearest = 0
for i in range(len(ar)):
dist = jdist(p, ar[i])
if (dist < mdist):
mdist = dist
nearest = i
return(nearest, mdist)
# -----------------------------------------------
path = "C:\\Users\\beale\\Documents\\Umap\\raw2" # raw input images here
img_count = 10000 # how many images to consider
px_count = 53*20 # how many pixels in each image (x * y)
dt = np.dtype(('U', 25)) # string with filename minus extension
sns.set(style='white', context='poster', rc={'figure.figsize': (14, 10)})
np.random.seed(17)
# data = np.random.rand(800, 768) # 1st var: items 2nd var: parts of item
data = np.zeros((img_count, px_count), dtype=np.float32) # items,parts of item
pname = np.ndarray(img_count, dt)
# Loop over (img_count) input image files and add them into data[]
# also add the filenames to pname[]
i = 0 # item index
for iname in os.listdir(path):
if (iname.startswith("DH5_")) and (iname.lower().endswith(".png")):
fname_in = os.path.join(path, iname)
print(i, fname_in)
pname[i] = (iname[0:25])
img = img_as_float(skimage.io.imread(fname=fname_in)) # img input
data[i, :] = img.reshape(px_count)
i = i + 1
if (i >= img_count):
break
# --- Generate the clustered similarity map for the data
# mapper = umap.UMAP(n_neighbors=30, min_dist=0.05).fit(data)
mapper = umap.UMAP(densmap=True, dens_lambda=1,
n_neighbors=10, min_dist=0.1).fit(data)
# mapper = umap.UMAP(n_neighbors=10, min_dist=0).fit(data)
umap.plot.diagnostic(mapper, diagnostic_type='vq')
umap.plot.points(mapper)
umap.plot.connectivity(mapper, show_points=True) # inter-connection lines
umap.plot.diagnostic(mapper, diagnostic_type='pca')
local_dims = umap.plot.diagnostic(mapper, diagnostic_type='local_dim')
umap.plot.diagnostic(mapper, diagnostic_type='neighborhood')
# display plot on an interactive web page
hover_data = pd.DataFrame({'index': np.arange(img_count),
'label': pname[:img_count]})
p = umap.plot.interactive(mapper, labels=pname[:img_count],
hover_data=hover_data, point_size=3)
umap.plot.show(p)
"""
i_cent = 8888 # index of item in center of ROI
i_edge = 8612 # index of item at edge of ROI
d_center = mapper.embedding_[i_cent] # coords of data at center ROI
r_active = jdist(d_center, mapper.embedding_[i_edge]) # radius of ROI
j = 0
for i in range(len(mapper.embedding_)):
r_test = jdist(d_center, mapper.embedding_[i])
if (r_test < r_active):
print(j, i, r_test, pname[i])
j += 1
n_ROI = j # count of data items within ROI
data1 = np.zeros((n_ROI, px_count), dtype=np.float32) # ROI only
# imap1 = np.zeros(n_ROI, dtype=np.int32) # map new index to original index
pname1 = np.ndarray(n_ROI, dt) # filenames in ROI
j = 0
for i in range(len(mapper.embedding_)):
r_test = jdist(d_center, mapper.embedding_[i])
if (r_test < r_active):
data1[j, :] = data[i, :]
pname1[j] = pname[i] # save the ROI name from original names
j += 1
# mapper1 = umap.UMAP(n_neighbors=10, min_dist=0.05).fit(data1)
# mapper1 = umap.UMAP(n_neighbors=10, min_dist=0.02).fit(data1)
mapper1 = umap.UMAP(densmap=True, dens_lambda=4,
n_neighbors=10, min_dist=0.2).fit(data1)
umap.plot.diagnostic(mapper1, diagnostic_type='vq')
"""
"""
# reducer = umap.UMAP(n_neighbors=30, min_dist=0.10)
# embedding = reducer.fit_transform(data)
# import joblib
# save_filename = 'Car25k-reduc1.sav'
# joblib.dump(reducer, save_filename)
# loaded_reducer = joblib.load(save_filename)
"""
"""
hover_data1 = pd.DataFrame({'index': np.arange(n_ROI),
'label': pname1})
p1 = umap.plot.interactive(mapper1, labels=pname1,
hover_data=hover_data1, point_size=3)
umap.plot.show(p1)
"""
"""
# fit = umap.UMAP(n_neighbors=60, min_dist = 0.15)
# u = fit.fit_transform(data)
# plt.scatter(u[:,0], u[:,1], c=data[:,0:4])
# plt.scatter(u[:, 0], u[:, 1])
# plt.title('UMAP embedding of car images')
"""
"""
reducer = umap.UMAP(n_neighbors=30, min_dist=0.10)
embedding = reducer.fit_transform(data)
import joblib
save_filename = 'CarB10kB-reduce.sav'
joblib.dump(reducer, save_filename)
fout = "G2-cars10k.txt"
with open(fout, 'w') as f:
f.write("idx,x,y,fname\n")
for j in range(len(data)):
f.write("%d, %5.3f,%5.3f, %s.jpg\n" % (j, mapper.embedding_[j, 0],
mapper.embedding_[j,1], pname[j]))
f.close()
fout = "G2-cars3636.csv"
with open(fout, 'w') as f:
f.write("idx,x,y,fname\n")
for j in range(n_ROI):
f.write("%d, %5.3f,%5.3f, %s.jpg\n" % (j, mapper1.embedding_[j, 0],
mapper1.embedding_[j,1], pname1[j]))
f.close()
"""
"""
# find data element nearest each to each regularly-spaced grid vertex
p = np.array([1, 1])
xsteps = 10
ysteps = 20
xa = -3 # min..max X range on UMAP axis
xb = 12
ya = -4 # min..max Y range on UMAP axis
yb = 14
xs = (xb - xa) / xsteps
ys = (yb - ya) / ysteps
print("x,y,n,dist,fname")
for x in range(xsteps):
for y in range(ysteps):
xp = xa + x*xs
yp = ya + y*ys
p = np.array([xp, yp])
(index, dist) = jmin(p, mapper1.embedding_)
print("%d,%d,%d,%5.2f,%s" %
(x, y, index, dist, pname1[index]), end='\n')
"""