-
Notifications
You must be signed in to change notification settings - Fork 0
/
plot_eval.py
174 lines (143 loc) · 5.51 KB
/
plot_eval.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
# from mpl_toolkits.mplot3d import axes3d
import numpy as np
import matplotlib.pyplot as plt
import IPython as ip
plt.rc('text', usetex=True)
plt.rc('font', family='serif')
plt.style.use('seaborn-paper')
# avoid to use Type 3 font which violates IEEE PaperPlaza
import matplotlib
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
# plt.style.use('ggplot')
# plt.style.use('default')
# plt.style.use('seaborn-whitegrid')
# plt.style.use('presentation')
# fmt_list = ['-', '--', '-.']
fmt_list = ['-', '-', '-']
# CHANGE THIS LIST TO DISABLE A SEPECIFIC METHOD
ignore_list = ['fcn']
# ignore_list = []
def plot_error_bar(x, m, s, gms, title='', enable_legend=False, xlabel='', ylabel=''):
fig, ax = plt.subplots(figsize=(4, 3))
for i, gm in enumerate(gms):
if gm[1] in ignore_list:
continue
# ax.errorbar(x, m[:, i], yerr=s[:, i], label=gm[1].upper(), fmt=fmt_list[i], linewidth=2)
ax.plot(x, m[:, i], label=gm[1].upper())
ax.fill_between(x, m[:, i]-s[:, i], m[:, i]+s[:, i], alpha=0.4)
if enable_legend:
ax.legend(loc='upper right')
ax.set_title(title)
ax.set_ylim(0, 1.03)
ax.set_xlabel(xlabel)
ax.set_ylabel(ylabel)
ax.set_xticks(x)
fig.tight_layout()
# fig.subplots_adjust(bottom=0.15) # for xlabel space
return fig
from matplotlib import cm
def plot_3d(x, y, m, title='', xlabel='', ylabel=''):
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
# for color values refer to https://matplotlib.org/users/dflt_style_changes.html
color_list = ['#1f77b4', '#ff7f0e', '#2ca02c']
for i in range(m.shape[2]):
Z = m[:, :, i]
X = np.empty_like(Z)
Y = np.empty_like(Z)
for j in range(X.shape[1]):
X[:, j] = x
for k in range(Y.shape[0]):
Y[k, :] = y
# ax.plot_wireframe(X, Y, Z, rstride=1, cstride=1, cmap=cm.jet)
ax.plot_surface(X, Y, Z, rstride=1, cstride=1, color=color_list[i], alpha=0.5, edgecolors=color_list[i])
# ax.axis('equal')
ax.set_title(title)
ax.set_zlim(0, 1)
ax.set_xlabel(xlabel)
ax.set_ylabel(ylabel)
ax.set_zlabel('accuracy')
ax.set_xticks(x)
ax.set_yticks(y)
return fig
def generate_table(m, s, name):
str_names = ''
str_acc = name + ' accuracy (\%) '
for i, gm in enumerate(gms):
str_names += '& \\{} '.format(gm[1].upper())
str_acc += '& ${0:.2f}\pm{1:.2f}$ '.format(m[0, i]*100, s[0, i]*100)
str_names += '\\\\'
str_acc += '\\\\'
return str_names + '\n' + str_acc
def plot_eval_results(acc_array, num_trials, range_nv, range_os, gms, num_interests=1, save=False):
# title_list = ['Wrist orientation', 'Grasping direction']
title_list = [r'\textbf{Wrist orientation} $\hat{\omega}$', r'\textbf{Grasping direction} $\hat{\delta}$']
interest_list = ['wrist', 'appr']
for i in range(num_interests):
# plot acc vs noise level
x = range_nv
if num_interests == 1:
y = acc_array[:, 0, :, :]
else:
y = acc_array[i, :, 0, :, :]
m = np.mean(y, axis=2)
s = np.std(y, axis=2)
f = plot_error_bar(x, m, s, gms,
title=title_list[i],
xlabel=r'\textbf{Number of noise voxels}', ylabel=r'\textbf{Accuracy}')
f.show()
if save:
f.savefig(interest_list[i]+'_nv.pdf', format='pdf', bbox_inches='tight')
# plot acc vs occlusion level
x = range_os
if num_interests == 1:
y = acc_array[0, :, :, :]
else:
y = acc_array[i, 0, :, :, :]
m = np.mean(y, axis=2)
s = np.std(y, axis=2)
# plot_error_bar(x[:-2], m[:-2], s[:-2], gms,
f = plot_error_bar(x[:], m[:], s[:], gms,
title=title_list[i],
enable_legend=True if i == 0 else False, xlabel=r'\textbf{Number of occluded voxel planes}', ylabel=r'\textbf{Accuracy}')
f.show()
if save:
f.savefig(interest_list[i]+'_ov.pdf', format='pdf', bbox_inches='tight')
# # plot acc vs both noise and occlusion level (3D plot)
# x = range_nv
# y = range_os
# if num_interests == 1:
# m = np.mean(acc_array, axis=3)
# else:
# m = np.mean(acc_array[i], axis=3)
# plot_3d(x, y, m, xlabel='Number of noise voxels', ylabel='Number of occluded voxel planes', title=title_list[i]).show()
# generate tables
print(generate_table(m, s, interest_list[i]))
# d = np.load('eval_grasping_1496631000.34.npz')
# d = np.load('eval_grasping_1496684052.51.npz')
# d = np.load('eval_grasping_1496711275.29.npz')
# d = np.load('eval_grasping_1496793016.78.npz')
# wrist & appr
# d = np.load('eval_grasping_1496935032.77.npz')
# d = np.load('eval_grasping_1496942069.3.npz')
# d = np.load('eval_grasping_1497001018.1.npz')
# d = np.load('eval_grasping_1497360808.86.npz')
# with banana (not bat)
# d = np.load('eval_grasping_1497538512.51.npz')
# with fcn and random as well
# d = np.load('eval_grasping_1497571857.33.npz') # very basic eval
# d = np.load('eval_grasping_1497635401.17.npz')
d = np.load('eval_grasping_1497801952.25.npz')
acc_array = d['acc_array']
try:
num_interests = d['num_interests']
except KeyError:
print('num_interests is not available, force to one')
num_interests = 1
num_trials = d['num_trials']
range_nv = d['range_nv']
range_os = d['range_os']
gms = d['gms']
plot_eval_results(acc_array, num_trials, range_nv, range_os, gms, num_interests=num_interests)
ip.embed()