-
Notifications
You must be signed in to change notification settings - Fork 9
/
metrics.py
219 lines (175 loc) · 6.54 KB
/
metrics.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
import numpy as np
import pandas as pd
import utils as ut
import sys
import os
import sklearn.metrics as m
import visualization as vis
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
def get_evaluation(path):
f = ut.get_latest_file(path, filter=r'.*.npy$')
print(f)
dict = np.load(f).item()
return dict
def distance(ref, pred):
error = ref - pred
return l2(error)
def l2(error):
error = error ** 2
error = np.sum(error, axis=1)
return np.sqrt(error)
def distance_improvement(prediction_dist, naive_dist):
pred_mean, naive_mean = np.mean(prediction_dist), np.mean(naive_dist)
improvement = naive_mean-pred_mean if naive_mean-pred_mean > 0 else 0
improvement = improvement / naive_mean
# print('predictive error(naive): %.9f (%.9f) -> %.2f%%'
# % (pred_mean, naive_mean, improvement*100))
return improvement, pred_mean, naive_mean
def distance_binary_improvement(prediction_dist, naive_dist):
pairwise = naive_dist - prediction_dist
pairwise[pairwise > 0] = 1
pairwise[pairwise < 0] = 0
fraction = np.mean(pairwise)
# print('Pairwise error improved for : %f%%' % (fraction*100), pairwise)
return fraction
def nn_metric(point_array):
"""
For every frame how often previous frame t-1 and next frame t+1 are within top-2 nearest neighbours
:param point_array: point array
:return:
"""
total = 0
for i in range(1, len(point_array)-1):
x = point_array - point_array[i]
d = l2(x)
indexes = np.argsort(d)[:3]
# assert i in indexes or d[indexes[0]][0] == 0
if i-1 in indexes:
total += 1
if i+1 in indexes:
total += 1
# print(i, i-1 in indexes, i+1 in indexes, indexes)
metric = total/(len(point_array) - 2) / 2
# print('NN metric: %.7f%%' % (metric * 100))
return metric
def nn_metric_pred(prediction, target):
"""
For every frame how often previous frame t-1 and next frame t+1 are within top-2 nearest neighbours
:param target: point array
:return:
"""
total = 0
for i in range(len(target)):
x = target - prediction[i]
d = l2(x)
index = np.argmin(d)
# print(index, i, d)
if i == index:
total += 1
metric = total/len(target)
# print('NN metric for preditcion: %.7f%%' % (metric * 100))
return metric
def test_nn():
enc = np.arange(0, 100).reshape((100, 1))
# enc.transpose()
# print(enc.shape)
# print(nn_metric(enc))
assert nn_metric(enc) == 1.
def test_nn_pred():
enc = np.arange(0, 100).reshape((100, 1))
# enc.transpose()
# print(enc.shape)
# print(nn_metric_pred(enc, enc+0.2))
assert nn_metric_pred(enc, enc) == 1.
def reco_error(x, y):
delta = x-y
error = m.mean_squared_error(x.flatten(), y.flatten())
return error
def print_folder_metrics(path):
eval = get_evaluation(path)
enc = eval['enc']
pred = enc[1:-1]*2 - enc[0:-2]
ref = enc[2:]
# print(enc[0], enc[1], pred[0], enc[2])
pred_to_target_dist, next_dist = distance(ref, pred), distance(enc[1:-1], enc[2:])
pl2, pred_d, naiv_d = distance_improvement(pred_to_target_dist, next_dist)
pb = distance_binary_improvement(pred_to_target_dist, next_dist)
pnn = nn_metric(enc)
pnnp = (nn_metric_pred(pred, ref)*100)
lreco = reco_error(eval['rec'], eval['blu'])
info = '%.3f & %.3f & %.3f & %.2f' % (pl2, pb, pnn, lreco)
print(info)
print('pimp:%f(%f/%f) & pb:%f & pnn:%f' % (pl2, pred_d, naiv_d, pb, pnn), 'pnnp: %f' % pnnp)
return info
def plot_single_cross_section_3d(data, select, subplot):
data = data[:, select]
# subplot.scatter(data[:, 0], data[:, 1], s=20, lw=0, edgecolors='none', alpha=1.0,
# subplot.plot(data[:, 0], data[:, 1], data[:, 2], color='black', lw=1, alpha=0.4)
d = data
# subplot.plot(d[[-1, 0], 0], d[[-1, 0], 1], d[[-1, 0], 2], lw=1, alpha=0.8, color='red')
# subplot.scatter(d[[-1, 0], 0], d[[-1, 0], 1], d[[-1, 0], 2], lw=10, alpha=0.3, marker=".", color='b')
d = data
subplot.scatter(d[:, 0], d[:, 1], d[:, 2], s=4, alpha=1.0, lw=0.5,
c=vis._build_radial_colors(len(d)),
marker=".",
cmap=plt.cm.hsv)
subplot.plot(data[:, 0], data[:, 1], data[:, 2], color='black', lw=0.2, alpha=0.9)
subplot.set_xlim([-0.01, 1.01])
subplot.set_ylim([-0.01, 1.01])
subplot.set_zlim([-0.01, 1.01])
ticks = []
subplot.xaxis.set_ticks(ticks)
subplot.yaxis.set_ticks(ticks)
subplot.zaxis.set_ticks(ticks)
subplot.xaxis.set_major_formatter(ticker.FormatStrFormatter('%1.0f'))
subplot.yaxis.set_major_formatter(ticker.FormatStrFormatter('%1.0f'))
def plot_single_cross_section_line(data, select, subplot):
data = data[:, select]
# subplot.scatter(data[:, 0], data[:, 1], s=20, lw=0, edgecolors='none', alpha=1.0,
# subplot.plot(data[:, 0], data[:, 1], data[:, 2], color='black', lw=1, alpha=0.4)
d = data
# subplot.plot(d[[-1, 0], 0], d[[-1, 0], 1], d[[-1, 0], 2], lw=1, alpha=0.8, color='red')
# subplot.scatter(d[[-1, 0], 0], d[[-1, 0], 1], d[[-1, 0], 2], lw=10, alpha=0.3, marker=".", color='b')
d = data
subplot.plot(data[:, 0], data[:, 1], data[:, 2], color='black', lw=1, alpha=0.4)
subplot.set_xlim([-0.01, 1.01])
subplot.set_ylim([-0.01, 1.01])
subplot.set_zlim([-0.01, 1.01])
ticks = []
subplot.xaxis.set_ticks(ticks)
subplot.yaxis.set_ticks(ticks)
subplot.zaxis.set_ticks(ticks)
subplot.xaxis.set_major_formatter(ticker.FormatStrFormatter('%1.0f'))
subplot.yaxis.set_major_formatter(ticker.FormatStrFormatter('%1.0f'))
if __name__ == '__main__':
path = os.getcwd()
for _, paths, _ in os.walk(path):
print('dirs', paths)
break
if len(paths) == 0:
print_folder_metrics(path)
eval = get_evaluation(path)
enc = eval['enc']
fig = vis.get_figure(shape=[1800, 900, 3])
# ax =
plot_single_cross_section_3d(enc, [0, 1, 2], plt.subplot(121, projection='3d'))
plot_single_cross_section_line(enc, [0, 1, 2], plt.subplot(122, projection='3d'))
plt.tight_layout()
plt.show()
else:
res = []
for d in paths + [path]:
print(d)
c_path = os.path.join(path, d)
info = print_folder_metrics(c_path)
res.append('\n%30s:\n%s' % (d, info))
res = sorted(res)
print('\n'.join(res))
print(len(paths), len(res))
exit(0)
if 'TensorFlow_DCIGN' in os.getcwd().split('/')[-1]:
# path = '/home/eugene/repo/TensorFlow_DCIGN/tmp/noise.f20_f4__i_grid03.14.c'
# path = '/mnt/code/vd/TensorFlow_DCIGN/tmp/pred.f101_f3__i_romb8.5.6'
path = '/media/eugene/back up/VD_backup/tmp_epoch20_final/pred.16c3s2_32c3s2_32c3s2_23c3_f3__i_romb8.5.6_'
# path = '/media/eugene/back up/VD_backup/tmp_epoch19_inputs/pred.16c3s2_32c3s2_32c3s2_16c3_f100_f3__i_grid.28.gh.360'