-
Notifications
You must be signed in to change notification settings - Fork 0
/
visualization.py
334 lines (270 loc) · 11.1 KB
/
visualization.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
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
import numpy as np
import itertools
import csv
import torch
from matplotlib import pyplot as plt
import os
from metrics import tensor_OSRC
def get_colors():
colors = np.array([
[230, 25, 75],
[60, 180, 75],
[255, 225, 25],
[67, 99, 216],
[245, 130, 49],
[145, 30, 180],
[70, 240, 240],
[240, 50, 230],
[188, 246, 12],
[250, 190, 190],
[0, 128, 128],
[230, 190, 255],
[154, 99, 36],
[255, 250, 200],
[128, 0, 0],
[170, 255, 195],
[128, 128, 0],
[255, 216, 177],
[0, 0, 117],
[128, 128, 128],
[255, 255, 255],
[0, 0, 0]
]).astype(np.float)
colors = colors / 255.
return colors
def simplescatter(features, classes, eps=None, eps_iter=None, current_iteration=None,
c=("b", "g", "r", "c", "m", "y", "orange", "lawngreen", "peru", "deeppink", "k"),
s=0.1):
plt.figure(1)
# scatterplot every digit to a color
for i in range(classes):
plt.scatter(*zip(*(features[i])), c=c[i], s=s)
plt.legend(range(classes))
if not os.path.exists('./plots'):
os.makedirs('./plots')
if eps and eps_iter and current_iteration:
plt.savefig(f"plots/flower_{eps}eps_{eps_iter}epsiter_{current_iteration}iter.png", dpi=600)
else:
plt.savefig("plots/flower.png", dpi=600)
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()
def epsilon_plot(eps_tensor, eps_list, eps_iter_list, title, iteration=None):
plt.figure(2)
# pull out the 3rd (depth) dimension of the tensor. Now for every eps-eps_iter pair theres a list with
# confidences over all epochs
confidences = eps_tensor.reshape(len(eps_tensor), -1).transpose(0, 1)
max_conf = 0
for i in range(len(confidences)):
eps_index = i // len(eps_iter_list)
eps_iter_index = i % len(eps_iter_list)
if eps_list[eps_index] == eps_iter_list[eps_iter_index]:
plt.plot(confidences[i], label=f"eps: {eps_list[eps_index]}")
plt.xlabel("epochs")
plt.ylabel(title)
plt.legend()
if not os.path.exists('./plots'):
os.makedirs('./plots')
if iteration:
plt.savefig(f"plots/epsilons_iter{iteration}_{title}.png")
else:
plt.savefig(f"plots/epsilons_{title}.png")
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()
def epsilon_table(eps_tensor, eps_list, eps_iter_list, title, iteration=None):
plt.figure(figsize=(6, 4), dpi=800)
data = eps_tensor[-1].cpu().detach().numpy()
max_idx = np.unravel_index(data.argmax(), data.shape)
cell_colors = np.full(data.shape, "w")
cell_colors[max_idx[0]][max_idx[1]] = "g"
columns = eps_iter_list
rows = eps_list
plt.axis('tight')
plt.axis('off')
plt.title(title)
plt.table(cellText=data, rowLabels=rows, colLabels=columns, loc="center", cellColours=cell_colors)
if not os.path.exists('./plots'):
os.makedirs('./plots')
if iteration:
plt.savefig(f"plots/table_iter{iteration}_{title}.png")
else:
plt.savefig(f"plots/table_{title}.png")
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()
def plot_histogram(pos_features, neg_features, pos_labels='Knowns', neg_labels='Unknowns', title="Histogram",
file_name='{}foo.pdf'):
"""
This function plots the Histogram for Magnitudes of feature vectors.
implementation taken from https://github.com/Vastlab/vast
"""
pos_mag = np.sqrt(np.sum(np.square(pos_features), axis=1))
neg_mag = np.sqrt(np.sum(np.square(neg_features), axis=1))
pos_hist = np.histogram(pos_mag, bins=500)
neg_hist = np.histogram(neg_mag, bins=500)
fig, ax = plt.subplots(figsize=(4.5, 1.75))
ax.plot(pos_hist[1][1:], pos_hist[0].astype(np.float16) / max(pos_hist[0]), label=pos_labels, color='g')
ax.plot(neg_hist[1][1:], neg_hist[0].astype(np.float16) / max(neg_hist[0]), color='r', label=neg_labels)
ax.tick_params(axis='both', which='major', labelsize=12)
plt.xscale('log')
plt.tight_layout()
if title is not None:
plt.title(title)
plt.savefig(file_name.format('Hist', 'pdf'), bbox_inches='tight')
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()
def plotter_2D(
pos_features,
labels,
neg_features=None,
pos_labels='Knowns',
neg_labels='Unknowns',
title=None,
file_name='foo.pdf',
final=False,
pred_weights=None,
heat_map=False):
colors = get_colors()
plt.figure(figsize=[6, 6])
"""
implementation taken from https://github.com/Vastlab/vast
"""
if heat_map:
min_x, max_x = np.min(pos_features[:, 0]), np.max(pos_features[:, 0])
min_y, max_y = np.min(pos_features[:, 1]), np.max(pos_features[:, 1])
x = np.linspace(min_x * 1.5, max_x * 1.5, 200)
y = np.linspace(min_y * 1.5, max_y * 1.5, 200)
pnts = list(itertools.chain(itertools.product(x, y)))
pnts = np.array(pnts)
e_ = np.exp(np.dot(pnts, pred_weights))
e_ = e_ / np.sum(e_, axis=1)[:, None]
res = np.max(e_, axis=1)
plt.pcolor(x, y, np.array(res).reshape(200, 200).transpose(), rasterized=True)
if neg_features is not None:
# Remove black color from knowns
colors = colors[:-1, :]
colors_with_repetition = colors.tolist()
for i in range(int(len(set(labels.tolist())) / colors.shape[0])):
colors_with_repetition.extend(colors.tolist())
colors_with_repetition.extend(colors.tolist()[:int(colors.shape[0] % len(set(labels.tolist())))])
colors_with_repetition = np.array(colors_with_repetition)
plt.scatter(pos_features[:, 0], pos_features[:, 1], c=colors_with_repetition[labels.astype(np.int)],
edgecolors='none', s=0.5)
if neg_features is not None:
plt.scatter(neg_features[:, 0], neg_features[:, 1], c='k', edgecolors='none', s=15, marker="*")
if final:
plt.gca().spines['right'].set_position('zero')
plt.gca().spines['bottom'].set_position('zero')
plt.gca().spines['left'].set_visible(False)
plt.gca().spines['top'].set_visible(False)
plt.tick_params(axis='both', bottom=False, left=False, labelbottom=False, labeltop=False, labelleft=False,
labelright=False)
plt.axis('equal')
plt.savefig(file_name.format('2D_plot', 'png'), bbox_inches='tight')
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()
if neg_features is not None:
plot_histogram(pos_features, neg_features, pos_labels=pos_labels, neg_labels=neg_labels, title=title,
file_name=file_name.format('hist', 'pdf'))
def sigmoid_2D_plotter(
pos_features,
labels,
neg_features=None,
pos_labels='Knowns',
neg_labels='Unknowns',
title=None,
file_name='foo.pdf',
final=False,
pred_weights=None,
heat_map=False):
"""
implementation taken from https://github.com/Vastlab/vast
"""
colors = get_colors()
plt.figure(figsize=[6, 6])
if heat_map:
min_x, max_x = np.min(pos_features[:, 0]), np.max(pos_features[:, 0])
min_y, max_y = np.min(pos_features[:, 1]), np.max(pos_features[:, 1])
x = np.linspace(min_x * 1.5, max_x * 1.5, 200)
y = np.linspace(min_y * 1.5, max_y * 1.5, 200)
pnts = list(itertools.chain(itertools.product(x, y)))
pnts = np.array(pnts)
e_ = np.exp(np.dot(pnts, pred_weights))
e_ = e_ / np.sum(e_, axis=1)[:, None]
res = np.max(e_, axis=1)
plt.pcolor(x, y, np.array(res).reshape(200, 200).transpose(), rasterized=True)
if neg_features is not None:
# Remove black color from knowns
colors = colors[:-1, :]
colors_with_repetition = colors.tolist()
for i in range(10):
plt.scatter(pos_features[labels == i, 0], pos_features[labels == i, 1], c=colors_with_repetition[i],
edgecolors='none', s=1. - (i / 10))
if neg_features is not None:
plt.scatter(neg_features[:, 0], neg_features[:, 1], c='k', edgecolors='none', s=15, marker="*")
if final:
plt.gca().spines['right'].set_position('zero')
plt.gca().spines['bottom'].set_position('zero')
plt.gca().spines['left'].set_visible(False)
plt.gca().spines['top'].set_visible(False)
plt.tick_params(axis='both', bottom=False, left=False, labelbottom=False, labeltop=False, labelleft=False,
labelright=False)
plt.axis('equal')
plt.savefig(file_name.format('2D_plot', 'png'), bbox_inches='tight')
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()
if neg_features is not None:
plot_histogram(pos_features, neg_features, pos_labels=pos_labels, neg_labels=neg_labels, title=title,
file_name=file_name.format('hist', 'pdf'))
def add_OSCR(name, to_plot=None):
"""
implementation taken from https://github.com/Vastlab/vast
"""
if to_plot is None:
to_plot = []
with open(f'models/mnist_scores.csv', mode='r') as file:
next(file)
mnist_data = list(csv.reader(file))
mnist_data = [x for x in mnist_data if x] # filter out empty lists
mnist_data = np.array(mnist_data, dtype=np.float32)
with open(f'models/letters_scores.csv', mode='r') as file:
next(file)
letters_data = list(csv.reader(file))
letters_data = [x for x in letters_data if x] # filter out empty lists
letters_data = np.array(letters_data, dtype=np.float32)
all_gt = torch.tensor(mnist_data[:, 0].tolist() + (-1 * torch.ones(letters_data.shape[0])).tolist()).view(-1)
all_prob = torch.tensor(mnist_data[:, 1:11].tolist() + letters_data[:, 1:11].tolist())
all_prob, all_predicted = torch.max(all_prob, dim=1)
knowns_accuracy, OSE = tensor_OSRC(all_gt, all_predicted, all_prob)
to_plot.append((knowns_accuracy, OSE, name.split('/')[-1]))
def plot_OSCR(to_plot, filename=None, title=None, no_of_false_positives=None):
"""
:param to_plot: list of tuples containing (knowns_accuracy,OSE,label_name)
:param no_of_false_positives: To write on the x axis
:param filename: filename to write
:return: None
implementation taken from https://github.com/Vastlab/vast
"""
fig, ax = plt.subplots()
if title is not None:
fig.suptitle(title, fontsize=20)
for plot_no, (knowns_accuracy, OSE, label_name) in enumerate(to_plot):
ax.plot(OSE, knowns_accuracy, label=label_name)
ax.set_xscale('log')
ax.autoscale(enable=True, axis='x', tight=True)
ax.set_ylim([0.8, 1])
ax.set_ylabel('Correct Classification Rate', fontsize=18, labelpad=10)
if no_of_false_positives is not None:
ax.set_xlabel(f"False Positive Rate : Total Unknowns {no_of_false_positives}", fontsize=18, labelpad=10)
else:
ax.set_xlabel(f"False Positive Rate", fontsize=18, labelpad=10)
ax.legend(loc="lower right")
if filename is not None:
fig.savefig(f"plots/{filename}.pdf", bbox_inches="tight")
if os.environ.get('PLOT') == "t":
plt.show()
plt.close()