forked from M-Lin-DM/Uniform-Orientation-Sampling
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Plotting_tools.py
162 lines (130 loc) · 6.08 KB
/
Plotting_tools.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
import plotly.graph_objs as go
import matplotlib.pyplot as plt
import numpy as np
# from scipy.spatial import Delaunay
from sklearn.neighbors import NearestNeighbors, KDTree, DistanceMetric
def KNN_from_vectors(V, K):
nbrs = NearestNeighbors(n_neighbors=K + 1, algorithm='kd_tree').fit(V)
distances, indices = nbrs.kneighbors(V)
distances, indices = distances[:, 1:], indices[:, 1:]
return distances, indices
def add_sphere(ax, scale):
u = np.linspace(0, 2 * np.pi, 100)
v = np.linspace(0, np.pi, 100)
x = scale * np.outer(np.cos(u), np.sin(v))
y = scale * np.outer(np.sin(u), np.sin(v))
z = scale * np.outer(np.ones(np.size(u)), np.cos(v))
# Setting an edgecolor appears to solve the issue of the surface hiding the points behind it.
# However, you may want to do this if plotting the triangulation network on the surface
ax.plot_surface(x, y, z, alpha=0.04, edgecolor='k')
# def add_delaunay(ax, emb):
# tri = Delaunay(emb)
# edges = np.concatenate((tri.simplices[:, :2], tri.simplices[:, 1:3]), axis=0)
# for j in range(len(edges)):
# line = np.array([emb[edges[j, 0]], emb[edges[j, 1]]])
# ax.plot3D(line[:, 0], line[:, 1], line[:, 2],'-', markerfacecolor='black', linewidth=1, color='black')
def add_KNN_network(ax, emb, plot_cap):
distances, indices = KNN_from_vectors(emb, 4)
for j in range(len(indices)):
for neighbor in range(indices.shape[1]):
line = np.array([emb[j], emb[indices[j, neighbor]]])
if plot_cap and emb[j, 2] > 0:
ax.plot3D(line[:, 0], line[:, 1], line[:, 2], '-', linewidth=0.5, color='black')
elif plot_cap==False:
ax.plot3D(line[:, 0], line[:, 1], line[:, 2], '-', linewidth=0.5, color='black')
class NDScatter:
"""
Class for plotting result of UOS algorithm
Parameters
------------
emb: ndarray
Each row should represent a vector in emb.shape[1]-dimensional space
UOS: UniformOrientationSampling object
UniformOrientationSampling object. See Uniform_Orientation_Sampling.py
make_raster: bool
whether to plot matrix containing embedding when dimensions>3
save_path: str
where to save plots
"""
def __init__(self, emb, UOS, make_raster=False, save_path=None):
self.emb = emb
self.dimensions = UOS.dimensions
self.pop_size = UOS.pop_size
self.iterations = UOS.iterations
self.approach = UOS.approach
self.save_path = save_path
self.make_raster = make_raster
def scatter2D(self):
a = 7
fig = plt.figure(figsize=(1.7778 * a,
a)) # e.g. figsize=(4, 3) --> img saved has resolution (400, 300) width by height when using dpi='figure' in savefig
plt.scatter(self.emb[:, 0], self.emb[:, 1], s=20, label='emb',
edgecolors='k', marker='o', facecolors='none')
# plt.xlim([0.0, 1.0])
# plt.ylim([0.0, 1.0])
plt.title('Embedding')
plt.gca().set_aspect('equal')
# plt.legend(loc="lower right")
plt.show()
if self.save_path is not None:
fig.savefig(self.save_path + 'emb_' + self.approach + '_' + str(self.dimensions) + 'D_N' + str(
self.pop_size) + '.png', transparent=False, dpi='figure', bbox_inches=None)
def scatter3D_view(self):
if self.emb.shape[1] == 3:
fig = go.Figure(data=[go.Scatter3d(
name='training images',
x=self.emb[:, 0],
y=self.emb[:, 1],
z=self.emb[:, 2],
mode='markers',
marker=dict(
size=1.5, color='black', symbol='circle')
)])
fig.show()
if self.save_path is not None:
fig.write_image(
self.save_path + 'emb_' + self.approach + '_' + str(
self.dimensions) + 'D_N' + str(self.pop_size) + '.png')
def plot_raster(self):
plt.figure(figsize=(20, 7))
ax = plt.axes()
plt.pcolormesh(np.arange(0, self.dimensions, 1), np.arange(0, self.pop_size), self.emb, shading='nearest',
cmap='inferno')
plt.xticks(fontsize=15)
plt.yticks(fontsize=15)
ax.set_yscale('linear')
ax.set_xlabel('component', fontsize=15)
ax.set_ylabel('orientation vector', fontsize=15)
cbar = plt.colorbar(ax=ax)
cbar.set_label('value', fontsize=15)
plt.show()
def scatter3D_mpl(self, plot_cap=True):
a = 8 # hundred pixels
phi = 55
theta = 45
fig2 = plt.figure(figsize=(1.7778 * a, a))
ax = fig2.add_subplot(projection='3d')
add_sphere(ax, 0.995)
cap = self.emb[:, 2] > 0
marker_size = lambda popsize: 100/popsize**0.6
if plot_cap and self.pop_size>8:
ax.plot3D(self.emb[cap, 0], self.emb[cap, 1], self.emb[cap, 2], 'o', markerfacecolor='black',
markersize=marker_size(self.pop_size), color='black')
else:
ax.plot3D(self.emb[:, 0], self.emb[:, 1], self.emb[:, 2], 'o', markerfacecolor='black',
markersize=marker_size(self.pop_size), color='black')
# add_delaunay(ax, self.emb) #failed since some tri
# add_KNN_network(ax, self.emb, plot_cap)
ax.dist = 8
ax.view_init(phi, theta) # view_init(elev=None, azim=None)
ax.grid(False)
if self.save_path is not None:
fig2.savefig(self.save_path + 'emb_' + self.approach + '_' + str(self.dimensions) + 'D_N' + str(
self.pop_size) + '.png', transparent=False, dpi='figure', bbox_inches=None)
def ndscatter(self):
if self.emb.shape[1] == 2:
self.scatter2D()
elif self.emb.shape[1] == 3:
self.scatter3D_view()
elif self.dimensions > 3 and self.make_raster:
self.plot_raster()