-
Notifications
You must be signed in to change notification settings - Fork 2
/
generate_TEST__data_linear_array.py
209 lines (180 loc) · 10.6 KB
/
generate_TEST__data_linear_array.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
#from numba import jit, cuda
import numpy as np
import os
import matplotlib.pyplot as plt
import sfs
import scipy
import tqdm
import argparse
from data_lib import params_linear_2D
from data_lib import soundfield_generation as sg
#import scikit-image as skimage
#from scikit-image.metrics import structural_similarity as ssim
#from skimage.metrics import structural_similarity as ssim
#plt.rcParams.update({
#"text.usetex": True,
#"font.family": "sans-serif",
#"font.sans-serif": ["Helvetica"],
#'font.size': 20})
idxy = params_linear_2D.idx_lr_gd_y
idx_y = idxy[int(len(idxy)/2):]
print("Running TESTing set generation")
def plot_soundfield(cmap, P, n_f, selection, axis_label_size, tick_font_size, save_path=None, plot_ldspks=True, do_norm=True):
figure = plt.figure(figsize=(20, 20))
if do_norm:
print("np.shape(P) = {}, \n np.shape(P[:, n_f]) = {} ".format(np.shape(P), np.shape(P[:, n_f])))
#im = sfs.plot2d.amplitude(np.reshape(P[:, n_f], (params_linear_2D.N_sample, params_linear_2D.N_sample)), params_linear_2D.grid2D, xnorm=[0, 0, 0], cmap=cmap, vmin=-1.0, vmax=1.0, colorbar=False)
#im = sfs.plot2d.amplitude(P[:, n_f], params_linear_2D.grid2D, xnorm=[0, 0, 0], cmap=cmap, vmin=-1.0, vmax=1.0, colorbar=False)
else:
print("in else")
#im = sfs.plot2d.amplitude(np.reshape(P[:, n_f], (params_linear_2D.N_sample, params_linear_2D.N_sample)), params_linear_2D.grid2D, cmap=cmap, colorbar=False, vmin=P[:, n_f].min(), vmax=P[:, n_f].max(), xnorm=None)
#im = sfs.plot2d.amplitude(P[:, n_f], params_linear_2D.grid2D, xnorm=None, cmap=cmap, vmin=-1.0, vmax=1.0, colorbar=False)
if plot_ldspks:
sfs.plot2d.loudspeakers(params_linear_2D.array.x[selection], params_linear_2D.array.n[selection], a0=1, size=0.18)
plt.xlabel('$x [m]$', fontsize=axis_label_size), plt.ylabel('$y [m]$', fontsize=axis_label_size)
plt.tick_params(axis='both', which='major', labelsize=tick_font_size)
#cbar = plt.colorbar(im, fraction=0.046)
#cbar.ax.tick_params(labelsize=tick_font_size)
# cbar.set_label('$NRE~[\mathrm{dB}]$',fontsize=tick_font_size))
plt.tight_layout()
if save_path is not None:
#plt.savefig(save_path)
print("save_path")
plt.show()
#@jit(target_backend='cuda')
def main():
# Arguments parse
parser = argparse.ArgumentParser(description='Generate data for linear array setup')
parser.add_argument('--base_dir', type=str, help="Base Data Directory", default='/nas/home/ralessandri/thesis_project/dataset')
#parser.add_argument('--base_dir', type=str, help="Base Data Directory", default='C:/Users/rales/OneDrive/Desktop/POLIMI/TESI/dataset')
##parser.add_argument('--dataset_name', type=str, help="Base Data Directory", default='data_src_wideband_point_W_23_test')
parser.add_argument('--gt_soundfield', type=bool, help='compute ground truth soundfield', default=True)
parser.add_argument('--n_missing', type=int, help='number missing loudspeakers', default=0)
args = parser.parse_args()
propagate_filters = False
eval_points = False
control_points = True
PLOT_RESULTS = False
if control_points:
c_points_x = params_linear_2D.idx_cp_x2
c_points_y = params_linear_2D.idx_cp_y2
c_pointsx_y = c_points_x[int(len(c_points_x) / 2):]
dataset_path = '/nas/home/ralessandri/thesis_project/dataset/test'
#dataset_path = 'C:/Users/rales/OneDrive/Desktop/POLIMI/TESI/pressure_matching_deep_learning/dataset/linear_array'
# Setup
# Grid of points where we actually compute the soundfield
point = params_linear_2D.point
#N_pts = len(point)
N_pts = int((len(params_linear_2D.grid2D[0][0])))
grid = params_linear_2D.grid2D
# Secondary Sources Green function
green_function_sec_sources_path = 'green_function_sec_sources_nl_' + str(params_linear_2D.N_lspks) + '_r_' + str(
params_linear_2D.rangeX[0]) + '.npy'
if os.path.exists(os.path.join(dataset_path, green_function_sec_sources_path)):
G = np.load(os.path.join(dataset_path, green_function_sec_sources_path))
else:
G = np.zeros((N_pts, N_pts, params_linear_2D.N_lspks, params_linear_2D.N_freqs), dtype=complex)
for n_l in tqdm.tqdm(range(params_linear_2D.N_lspks)):
for n_f in range(len(params_linear_2D.f_axis)):
# hankel_factor_1 = params_linear_2D.wc[n_f] / params_linear_2D.c # , (params_linear_2D.N_lspks, 1)
# print("array_pos = {}\nhankel_factor_2 = point[np] - array_pos = {}".format(params_linear_2D.array_pos, point - params_linear_2D.array_pos))
# hankel_factor_2 = np.linalg.norm(grid - params_linear_2D.array_pos[n_l]) # , reps=(params_linear_2D.N_freqs, 1) # ).transpose()
k = params_linear_2D.wc[n_f] / params_linear_2D.c
r = np.linalg.norm(grid - params_linear_2D.array_pos[n_l])
# np.exp(-1j * k * r) / (4 * np.pi)
# Points, Speakers, Frequencies
# G[:, :, n_l, n_f] = (1j / 4) * scipy.special.hankel2(0, hankel_factor_1*hankel_factor_2)
into_G = np.exp(-1j * k * r) / (4 * np.pi)
# print("into_G = {}".format(np.shape(into_G)))
G[:, :, n_l, n_f] = into_G # / r # ! Sto volutamente non tenendo conto del decay!
print("G is saved, shape = --> {}".format(np.shape(G)))
np.save(os.path.join(dataset_path, green_function_sec_sources_path), G)
# Check if array in grid points are equal
for n_p in range(N_pts):
if np.sum(np.linalg.norm(point[n_p] - params_linear_2D.array_pos, axis=1) == 0) > 0:
print(str(n_p))
N_missing = args.n_missing
N_lspks = params_linear_2D.N_lspks - N_missing
if N_missing>0:
lspks_config_path = 'lspk_config_nl_'+str(params_linear_2D.N_lspks)+'_missing_'+str(N_missing)+'.npy'
lspk_config_path_global = os.path.join(dataset_path, 'setup', lspks_config_path)
if os.path.exist(lspk_config_path_global):
idx_missing = np.load(lspk_config_path_global)
print('Loaded existing mic config')
else:
idx_missing = np.random.choice(np.arange(params_linear_2D.N_lspks), size=N_missing, replace=False)
print("saved missing indexes, shape --> {}".format(np.shape(idx_missing)))
np.save(lspk_config_path_global, idx_missing)
theta_l = np.delete(params_linear_2D.theta_l, idx_missing)
G = np.delete(G, idx_missing, axis=1)
if eval_points:
P_gt = np.zeros((len(params_linear_2D.src_pos_trainT), int(len(params_linear_2D.idx_lr_gd_y)), int(len(params_linear_2D.idx_lr_gd_x)), params_linear_2D.N_freqs), dtype=complex)
elif control_points:
P_gt = np.zeros((len(params_linear_2D.src_pos_trainT), int(len(c_points_y)), int(len(c_points_x)), params_linear_2D.N_freqs), dtype=complex)
else:
P_gt = np.zeros((len(params_linear_2D.src_pos_trainT), N_pts, N_pts, params_linear_2D.N_freqs),dtype=complex) # 3204, 64
d_array = np.zeros((params_linear_2D.N_lspks, N_lspks, params_linear_2D.N_freqs), dtype=complex)
for n_s in tqdm.tqdm(range(len(params_linear_2D.src_pos_testT))):
xs = np.append(params_linear_2D.src_pos_testT[n_s], 2)
# Multiply filters for hergoltz density function
if propagate_filters:
P = np.zeros((N_pts, len(params_linear_2D.wc)), dtype=complex)
for n_p in range(N_pts):
P[n_p, :] = np.sum(G[n_p] * d_array[n_s], axis=0)
if args.gt_soundfield:
for n_f in range(params_linear_2D.N_freqs):
# Sources, Points, Frequencies
#P_gt[n_s, :, :, n_f] = (1j / 4) * scipy.special.hankel2(0, (params_linear_2D.wc[n_f] / params_linear_2D.c) * np.linalg.norm(point[:, :2] - xs))
if eval_points:
into_P_gt_eval = sfs.fd.source.point(params_linear_2D.wc[n_f], xs, grid) * np.linalg.norm(grid - xs)
for i in params_linear_2D.idx_lr_gd_x:
for j in idx_y:
into_P_gt_eval[-j, i] = 0
into_P_gt_eval_y = into_P_gt_eval[params_linear_2D.idx_lr_gd_y]
P_gt[n_s, :, :, n_f] = into_P_gt_eval_y[:,params_linear_2D.idx_lr_gd_x]
elif control_points:
into_P_gt_eval = sfs.fd.source.point(params_linear_2D.wc[n_f], xs, grid) * np.linalg.norm(grid - xs) # this last multiplication is not valid if I consider the decay!
for i in c_points_x:
for j in c_pointsx_y:
into_P_gt_eval[-j, i] = 0
into_P_gt_eval_y = into_P_gt_eval[c_points_y]
P_gt[n_s, :, :, n_f] = into_P_gt_eval_y[:, c_points_x]
else:
P_gt[n_s, :, :, n_f] = sfs.fd.source.point(params_linear_2D.wc[n_f], xs, grid) * np.linalg.norm(grid - xs)
if PLOT_RESULTS:
# Plot params
print("Plotting")
selection = np.ones_like(params_linear_2D.array_pos[:, 0])
selection = selection == 1 # ?
#n_s = 32
n_f = 41
print(str(params_linear_2D.f_axis[n_f]))
cmap = 'RdBu_r'
tick_font_size = 70
axis_label_size = 90
plot_soundfield(cmap, P_gt[n_s], n_f, selection, axis_label_size, tick_font_size, save_path=None, plot_ldspks=True,
do_norm=False)
if args.gt_soundfield:
if control_points:
print("P_gt saved shape --> {}".format(np.shape(P_gt)))
np.save(os.path.join(dataset_path, 'gt_soundfield_train_half_cp_double_train.npy'), P_gt)
else:
print("P_gt saved shape --> {}".format(np.shape(P_gt)))
np.save(os.path.join(dataset_path, 'gt_soundfield_train.npy'), P_gt)
#np.save(os.path.join(dataset_path, 'filters_config_nl_'+str(N_lspks)+'_missing_'+str(N_missing)+'.npy'))
# P_gt = [sources, control points, frequencies]
#print("x = {}\ny = {}".format(np.shape(P_gt[:, :, 0]), np.shape(P_gt[0, :, :])))
#plt.figure(figsize=(10,10))
#plt.plot(P_gt[:64, :, 0].T, P_gt[0, :, :], 'g*')
#plt.xlabel('$x [m]$'), plt.ylabel('$y [m]$')
#plt.xlim(-2, 2)
#plt.ylim(-2, 2)
##plt.setzlim(0, 4)
##plt.legend(['eval points', 'control points', 'loudspeakers', 'train sources', 'test sources'])
#plt.title("#2D")
#plt.show()
print("Everything Defined")
if __name__ == '__main__':
print("Main running")
main()
print("## End Generate TESTing ##")