-
Notifications
You must be signed in to change notification settings - Fork 3
/
SVD.py
141 lines (113 loc) · 5.09 KB
/
SVD.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
import torch
import numpy as np
import os
import pickle
import math
from mmdet.visualization.visualize import Visualize_cv
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
def mkdir(path):
if os.path.exists(path) == False:
os.makedirs(path)
def save_pickle(dir_name, file_name, data):
'''
:param file_path: ...
:param data:
:return:
'''
mkdir(dir_name)
with open(dir_name + file_name + '.pickle', 'wb') as f:
pickle.dump(data, f, protocol=pickle.HIGHEST_PROTOCOL)
def load_pickle(file_path):
with open(file_path + '.pickle', 'rb') as f:
data = pickle.load(f)
return data
class SVD(object):
def __init__(self):
self.height = 768
self.width = 1280
self.dist = math.sqrt(self.height * self.height / 4 + self.width * self.width / 4) / 10
self.center = np.array([(640 - 1) / 2, (640 - 1) / 2])
self.r_coord = np.linspace(0, 360, 360, endpoint=False)
self.r_coord_x = torch.tensor(np.cos(self.r_coord*2*math.pi/360)).view(-1, 1)
self.r_coord_y = torch.tensor((-1) *np.sin(self.r_coord*2*math.pi/360)).view(-1, 1)
self.r_coord_xy = torch.cat((self.r_coord_x, self.r_coord_y), dim=1).type(torch.float32).cuda()
self.cen = torch.tensor(np.repeat(self.center.reshape(1, -1), 360, 0)).type(
torch.float32).cuda()
def update_matrix(self, data):
self.mat = torch.cat((self.mat, data.transpose(1, 0)), dim=1)
def load_contour_component(self):
f_list = os.listdir('data_pickle_f_sbd_360_re_')
nf_list = os.listdir('data_pickle_nf_sbd_360_re_')
f_list = [i.rstrip('.pickle') for i in f_list]
nf_list = [i.rstrip('.pickle') for i in nf_list]
# sampled
for i, name in enumerate(f_list):
data_p = load_pickle('data_pickle_f_sbd_360_re_/' + name)
if len(data_p) != 0:
self.update_matrix(data_p.cuda())
print('%d done!' % i)
for i, name in enumerate(nf_list):
data_p = load_pickle('data_pickle_nf_sbd_360_re_/' + name)
if len(data_p) != 0:
self.update_matrix(data_p.cuda())
print('%d done!' % i)
save_pickle(dir_name='data_pickle/',
file_name='matrix_re_',
data=self.mat)
def do_SVD(self):
self.mat = load_pickle('data_pickle/matrix_re')
n, l = self.mat.shape
idx = torch.linspace(0, l-1, 40000).type(torch.int64).cuda()
U, S, V = np.linalg.svd(self.mat.cpu().numpy() / (self.dist), full_matrices=True)
self.U = torch.from_numpy(U).cuda()
self.S = torch.from_numpy(S).cuda()
save_pickle(dir_name='data_pickle/',
file_name='U_re',
data=self.U)
save_pickle(dir_name='data_pickle/',
file_name='S_re',
data=self.S)
def make_dict(self):
self.mat = torch.FloatTensor([]).cuda()
self.U = torch.FloatTensor([]).cuda()
self.S = torch.FloatTensor([]).cuda()
self.V = torch.FloatTensor([]).cuda()
def visualization_U(self):
self.U = load_pickle("data_pickle/U")
for k in range(self.U.shape[1]):
if k == 6:
break
temp = np.full((640, 640, 3), fill_value=255, dtype=np.uint8)
self.visualize.show['candidates'] = np.copy(temp)
U = self.U[:, k:k + 1] * 3000
dc = torch.full((360, 1), fill_value=180).type(torch.float32).cuda()
xy = self.r_coord_xy * U
xy_dc = self.r_coord_xy * dc
polygon_pts = self.cen + xy
polygon_pts_dc = self.cen + xy_dc
allow_pts = torch.cat((self.cen, polygon_pts), dim=1)
# self.visualize.draw_arrowedlines_cv(data=to_np(allow_pts).astype(np.int64), name='candidates', interval=1, ref_name='candidates',
# color=(255, 255, 255), s=1)
self.visualize.draw_polyline_cv(data=polygon_pts.cpu().numpy(), name='candidates', ref_name='candidates',
color=(0, 0, 255), s=15)
# self.visualize.draw_polyline_cv(data=polygon_pts.numpy(), name='candidates', ref_name='candidates',
# color=(51, 153, 102), s=15)
# self.visualize.draw_points_cv(data=to_np(self.cen[0:1, :]), name='candidates', ref_name='candidates',
# color=(255, 255, 255), s=5)
dir_name = 'display_U_red/'
file_name = 'U_' + str(k + 1) + '.jpg'
# file_name = 'U_' + str(k) + '.png'
self.visualize.display_saveimg(dir_name=dir_name,
file_name=file_name,
list=['candidates'])
def run(self):
print('start')
self.make_dict()
self.visualize = Visualize_cv()
# self.load_contour_component()
# self.do_SVD()
self.visualization_U()
if __name__ == "__main__":
svd = SVD()
svd.run()