-
Notifications
You must be signed in to change notification settings - Fork 0
/
download.py
219 lines (169 loc) · 4.76 KB
/
download.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
209
210
211
212
213
214
215
216
217
218
219
from json import dump
from os import listdir, rename
from os.path import dirname, join, realpath, isdir, exists
import shutil
import DuckDuckGoImages as ddg
from duckduckgo_search import ddg_images
from matplotlib import pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
import numpy as np
import pandas as pd
from PIL import Image
from sklearn.cluster import MiniBatchKMeans
from sklearn.mixture import GaussianMixture
from sklearn.cluster import DBSCAN, OPTICS
_cdir = dirname(realpath(__file__))
_bdir = join(_cdir, "..")
dl_path = join(_cdir, 'dl_images')
tc = 32
term = "pikachu"
desc = "The poster child(?) of Pokemon"
name = "Pikachu (Pokemon)"
plot = False
save = not plot
download = False
remove_greys = True
grey_radius = 25
remove_whites = False
white_radius = 75
remove_blacks = False
black_radius = 75
double = False
triple = False
# Download images returned with a given search query
if download:
if exists(dl_path):
shutil.rmtree(dl_path)
# ddg.download(term, folder=dl_path, max_urls=50,
# thumbnails=True, remove_folder=True)
#
ddg_images(term, max_results=50, size="Small", download=True)
files = listdir(_cdir)
folder_path = [item for item in files if isdir(item)][0]
rename(folder_path, dl_path)
points = np.zeros((1, 1))
weights = np.zeros((1, 1))
for image_name in listdir(dl_path):
image_path = join(dl_path, image_name)
try:
image = Image.open(image_path)
except:
continue
pixels = np.asarray(image)[0::10, 0::10]
shape = pixels.shape
try:
w, h, l = shape
except:
w, h = shape
try:
pixels = pixels.reshape((w*h, 3))
except:
continue
mix = GaussianMixture(n_components=tc, random_state=0).fit(pixels)
if points.shape == (1, 1):
points = mix.means_
weights = mix.weights_
else:
points = np.append(points, mix.means_, axis=0)
weights = np.append(weights, mix.weights_, axis=0)
# points = points[weights > (1/64)]
# weights = weights[weights > (1/64)]
df = pd.DataFrame(points)
dbscan = OPTICS(min_samples=2, max_eps=6.5).fit(df)
df['label'] = dbscan.labels_
centers = pd.DataFrame(index=sorted(pd.unique(dbscan.labels_)))
centers['counts'] = df['label'].value_counts()
# if len(pd.unique(dbscan.labels_)) > 8:
# large = centers[centers['counts'] > 2]
# else:
# large = centers
means = df.groupby('label').mean()
means = means.iloc[centers.index]
means = means.loc[~means.index.isin([-1])]
if double:
dbscan = OPTICS(min_samples=5, max_eps=65).fit(means)
means['label2'] = dbscan.labels_
centers = pd.DataFrame(index=sorted(pd.unique(dbscan.labels_)))
centers['counts'] = means['label2'].value_counts()
means = means[means['label2'] != -1]
if triple:
dbscan = OPTICS(min_samples=5, max_eps=15).fit(means)
means['label3'] = dbscan.labels_
centers = pd.DataFrame(index=sorted(pd.unique(dbscan.labels_)))
centers['counts'] = means['label3'].value_counts()
means = means[means['label3'] != -1]
try:
del means['label']
except:
pass
try:
del means['label2']
except:
pass
try:
del means['label3']
except:
pass
points = means.values
actpoints = []
if remove_greys:
for point in points:
dists = [np.linalg.norm(point - [i, i, i]) for i in range(256)]
if min(dists) > grey_radius:
actpoints.append(point)
actpoints = np.array(actpoints)
else:
actpoints = points
newpoints = []
if remove_blacks:
for point in actpoints:
dist = np.linalg.norm(point - [0, 0, 0])
if dist > black_radius:
newpoints.append(point)
newpoints = np.array(newpoints)
else:
newpoints = actpoints
finpoints = []
if remove_whites:
for point in newpoints:
dist = np.linalg.norm(point - [255, 255, 255])
if dist > white_radius:
finpoints.append(point)
finpoints = np.array(finpoints)
else:
finpoints = newpoints
r = [point[0] for point in finpoints]
g = [point[1] for point in finpoints]
b = [point[2] for point in finpoints]
# print(counts)
if plot:
fig = plt.figure()
ax1=fig.add_subplot(111, projection='3d')
# r = [point[0] for point in points]
# g = [point[1] for point in points]
# b = [point[2] for point in points]
ax1.scatter(r, g, b, c=np.clip(finpoints/255, 0, 1), s=200, alpha=1, depthshade=False)
# r = [point[0] for point in pixels[0::10000]]
# g = [point[1] for point in pixels[0::10000]]
# b = [point[2] for point in pixels[0::10000]]
# ax1.scatter(r, g, b, c=pixels[0::10000]/255)
ax1.set_xlabel('R')
ax1.set_ylabel('G')
ax1.set_zlabel('B')
ax1.set_title(name)
ax1.view_init(27,200)
plt.show()
if save:
data = dict()
data['name'] = name
data['desc'] = desc
data['term'] = term
balls = []
for point in finpoints:
clipped = np.clip(point, 0, 255)
r, g, b = clipped
balls.append([int(r), int(g), int(b), 5])
data['balls'] = balls
filename = "{}.json".format(name.lower().replace(' ', '-'))
with open(join(_bdir, 'picker', 'themes', 'data', filename), 'w') as f:
dump(data, f)