Micha commited on
Commit
c5973f1
1 Parent(s): 0325897

Added generating code for simulations

Browse files
code/KwaveUpdateDomainParams.py ADDED
@@ -0,0 +1,274 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+ # coding: utf-8
3
+
4
+ import numpy as np
5
+ import h5py
6
+ import argparse
7
+ import random
8
+ import shutil
9
+ import os
10
+ import scipy.interpolate as si
11
+
12
+ from scipy.signal import gaussian
13
+ from itertools import product
14
+
15
+ from typing import Tuple
16
+
17
+ def generate_sound_speed(shape = [1152, 512],
18
+ c_ref: float = 1500., c_min: float = 1300., c_max: float = 1800.,
19
+ atn_min: float = 0.05, atn_max: float = 0.15,
20
+ max_ell: int = 6, max_lines: int = 4) -> Tuple[np.ndarray, np.ndarray]:
21
+ """[summary]
22
+ Generate a random speed of sound and attenuation map with up to max_ell ellipses and max_lines straight reflectors
23
+
24
+ Args:
25
+ shape (list, optional): [description]. Defaults to [1152, 512].
26
+ c_ref (float, optional): [description]. Defaults to 1500..
27
+ c_min (float, optional): [description]. Defaults to 1300..
28
+ c_max (float, optional): [description]. Defaults to 1800..
29
+ atn_min (float, optional): [description]. Defaults to 0.05.
30
+ atn_max (float, optional): [description]. Defaults to 0.15.
31
+ max_ell (int, optional): [description]. Defaults to 6.
32
+ max_lines (int, optional): [description]. Defaults to 4.
33
+
34
+ Returns:
35
+ [type]: [description]
36
+ """
37
+
38
+ while True:
39
+ n_ell = np.random.randint(max_ell + 1)
40
+ n_lines = np.random.randint(max_lines + 1)
41
+
42
+ if n_ell + n_lines > 0:
43
+ break
44
+
45
+ ell_type = np.array([0,] * n_ell + [1,] * n_lines)[np.random.permutation(n_ell + n_lines)]
46
+ c = np.random.rand(n_ell + n_lines + 1) * (c_max - c_min) + c_min
47
+ atn = np.random.rand(n_ell + n_lines + 1) * (atn_max - atn_min) + atn_min
48
+
49
+ ell_params = np.random.rand(5, n_ell); # x0, y0, a, b, \theta in columns
50
+
51
+ ell_params[0, :] = ell_params[0, :] * shape[-1]
52
+ ell_params[1, :] = ell_params[1, :] * shape[-2]
53
+ ell_params[2, :] = ell_params[2, :] * shape[-1];
54
+ ell_params[3, :] = ell_params[3, :] * shape[-2];
55
+ ell_params[4, :] = (ell_params[4, :] - 0.5) * np.pi / 4;
56
+
57
+ line_params = np.random.rand(2, n_lines); # Angle and offset
58
+
59
+ line_params[0, :] = (line_params[0, :] - 0.5) * np.pi / 6;
60
+ line_params[1, :] = line_params[1, :] * (shape[-2] - 68) + 68
61
+
62
+ sound_speed = np.empty(shape[-2:], dtype=np.single)
63
+ sound_speed[...] = c[-1]
64
+ alpha_coeff = np.empty(shape[-2:], dtype=np.single)
65
+ alpha_coeff[...] = atn[-1]
66
+
67
+ i_ellp = 0
68
+ i_lines = 0
69
+
70
+ X, Y = np.meshgrid(range(shape[-1]), range(shape[-2]))
71
+
72
+ for i,t in enumerate(ell_type):
73
+ if t == 0:
74
+ # elipse
75
+ x0 = np.array([ell_params[0, i_ellp], ell_params[1, i_ellp]])
76
+ a = np.array([ell_params[2, i_ellp], ell_params[3, i_ellp]])
77
+
78
+ th = ell_params[4, i_ellp]
79
+ ath = np.abs(th)
80
+
81
+ R = np.array([[np.cos(th), -np.sin(th)], [np.sin(th), np.cos(th)]])
82
+ aR = np.array([[np.cos(ath), -np.sin(ath)], [np.sin(ath), np.cos(ath)]])
83
+
84
+ toff = aR.dot(a)[1]
85
+
86
+ if x0[1] - toff < 68:
87
+ x0[1] = 68 + toff
88
+
89
+ x = X - x0[0]
90
+ y = Y - x0[1]
91
+
92
+ u = np.cos(th) * x - np.sin(th) * y
93
+ v = np.sin(th) * x + np.cos(th) * y
94
+
95
+ sound_speed[(u / ell_params[2, i_ellp]) ** 2 + (v / min(ell_params[3, i_ellp], ell_params[1, i_ellp] - 68)) ** 2 < 1] = c[i]
96
+ alpha_coeff[(u / ell_params[2, i_ellp]) ** 2 + (v / min(ell_params[3, i_ellp], ell_params[1, i_ellp] - 68)) ** 2 < 1] = atn[i]
97
+
98
+ i_ellp += 1
99
+ else:
100
+ # plane
101
+ th = line_params[0, i_lines]
102
+ off = line_params[1, i_lines]
103
+
104
+ if off - shape[-1] * np.tan(np.abs(th)) < 68:
105
+ off = shape[-1] * np.tan(np.abs(th)) + 68
106
+
107
+ sound_speed[X * np.sin(th) + (Y - off) * np.cos(th) > 0] = c[i]
108
+ alpha_coeff[X * np.sin(th) + (Y - off) * np.cos(th) > 0] = atn[i]
109
+
110
+ i_lines += 1
111
+
112
+ return sound_speed, alpha_coeff
113
+
114
+
115
+ def generate_density(shape: tuple = [1152, 1152],
116
+ d_ref: float = 900.,
117
+ n_spec: float = -1.,
118
+ spec_amp: float = 0.1) -> Tuple[np.ndarray, np.ndarray, np.ndarray]:
119
+ """
120
+ Generate a speckle map in the density domain, including the x and y derivatives
121
+
122
+
123
+ Args:
124
+ shape (_type_, optional): shape of output map. Defaults to [1152, 1152]
125
+ d_ref (float, optional): reference density (kg / m^3). Defaults to 900
126
+ n_spec (float, optional): speckle density. Defaults to -1
127
+ spec_amp (float, optional): speckle amplitude. Defaults to 0.1
128
+
129
+ Returns:
130
+ _type_: _description_
131
+ """
132
+
133
+ if n_spec < 0:
134
+ n_spec = shape[-1] * shape[-2] * (2. / 64.)
135
+
136
+ t_n_spec = int(np.round(n_spec * (10 - np.random.rand() * 8) * 0.1))
137
+
138
+ speckle = np.zeros((shape[-2] - 48 - 68, shape[-1] - 48*2), np.single)
139
+ speckle.reshape(-1,1)[np.random.permutation(speckle.size)[:t_n_spec]] = d_ref * (np.random.rand(t_n_spec, 1) - 1/3) * spec_amp
140
+
141
+ rho0 = np.empty(shape[-2:], dtype=np.single)
142
+ rho0[...] = d_ref
143
+
144
+ rho0[68:-48, 48:-48] += speckle
145
+
146
+ rho0_sgx = np.empty_like(rho0)
147
+ rho0_sgx[:,:-1] = (rho0[:,:-1] + rho0[:,1:]) * 0.5
148
+ rho0_sgx[:,-1] = rho0[:,-1]
149
+
150
+ rho0_sgy = np.empty_like(rho0)
151
+ rho0_sgy[:-1,:] = (rho0[:-1,:] + rho0[1:,:]) * 0.5
152
+ rho0_sgy[-1,:] = rho0[-1,:]
153
+
154
+ return rho0, rho0_sgx, rho0_sgy
155
+
156
+
157
+ def generate_pulse(center_freq=5e6, n_periods=4.4, sampling_freq=40e6):
158
+ samples_per_cycle = sampling_freq / center_freq
159
+ n_samples = np.ceil(samples_per_cycle * n_periods + 1)
160
+
161
+ signal = np.sin(np.arange(n_samples, dtype=np.float32) / samples_per_cycle * 2 * np.pi) * gaussian(n_samples, (n_samples - 1) / 6).astype(np.single)
162
+
163
+ return signal
164
+
165
+
166
+ if __name__ == "__main__":
167
+ parser = argparse.ArgumentParser()
168
+ parser.add_argument('input', help='reference file')
169
+ parser.add_argument('--device', '-d', type=int, default=0, help='cuda device to use')
170
+ parser.add_argument('--frequencies', '-f', type=float, nargs='+', default=[5e6, 2.5e6],
171
+ help='at which frequencies to run the simulation')
172
+ parser.add_argument('--offsets', '-o', type=int, nargs='+', default=[-32, -24, -16, -8, 0, 8, 16, 24, 32],
173
+ help='list of (signed) offsets to use with respect to center, in elements. Angles will be calculated from offsets')
174
+ parser.add_argument('--n_periods', '-p', type=float, default=5.0,
175
+ help='how many periods to use for the pulse')
176
+ parser.add_argument('--ref_offset', '-r', type=int, default=59 * 1152 + (1152 - (128 + 127) * 4) // 2 + 32 * 2 * 4,
177
+ help='reference offset in entries (where to start offset 0) - default: line 60, element 33, remember the kerf')
178
+ parser.add_argument('--ref_speed', '-s', type=float, default=1540,
179
+ help='reference speed under which to computer angle')
180
+
181
+ args = parser.parse_args()
182
+
183
+ hash = f'{random.getrandbits(128):x}'
184
+
185
+ tmp_input = f'tmp_input_{hash}.h5'
186
+ tmp_output = f'tmp_output_{hash}.h5'
187
+ sim_result = f'sim_result_{hash}.h5'
188
+
189
+ # costs an additional copy but solves problem of opening file multiple times in parallel
190
+ shutil.copyfile(args.input, tmp_input)
191
+
192
+ try:
193
+ with h5py.File(tmp_input, 'r+') as f_i:
194
+ shape = np.squeeze(f_i['c0'][()]).shape
195
+ dt = f_i['dt'][0,0,0]
196
+ dx = f_i['dx'][0,0,0]
197
+
198
+ c0, alpha_coeff = generate_sound_speed(shape)
199
+ rho0, rho0_sgx, rho0_sgy = generate_density(shape)
200
+
201
+ f_i['c0'][...] = c0
202
+ f_i['alpha_coeff'][...] = alpha_coeff
203
+
204
+ f_i['rho0'][...] = rho0
205
+ f_i['rho0_sgx'][...] = rho0_sgx
206
+ f_i['rho0_sgy'][...] = rho0_sgy
207
+
208
+ # # Also record receive traces
209
+ # smi = np.tile(np.array(f_i['sensor_mask_index']), (1,1,2))
210
+ # smi[...,smi.shape[-1]//2:] += 1152 * (1152 - 60)
211
+ # del f_i['sensor_mask_index']
212
+ # f_i["sensor_mask_index"] = smi
213
+ # f_i["sensor_mask_index"].attrs["data_type"] = np.string_(b"long")
214
+ # f_i["sensor_mask_index"].attrs["domain_type"] = np.string_(b"real")
215
+
216
+ with h5py.File(sim_result, 'w') as f_o:
217
+ f_o.create_dataset(f'c0', data=c0, compression="gzip", compression_opts=9)
218
+ f_o.create_dataset(f'alpha_coeff', data=alpha_coeff, compression="gzip", compression_opts=9)
219
+ f_o.create_dataset(f'f', data=np.array(args.frequencies))
220
+ f_o.create_dataset(f'offsets', data=np.array(args.offsets))
221
+
222
+ for frequency, offset in product(args.frequencies, args.offsets):
223
+ # TODO: deal with offsets / angles
224
+ w = 127 * 4
225
+ o = offset * 8
226
+ dn = w * o / np.sqrt(64 * o ** 2 + 4 * w ** 2) * (dx / dt) / args.ref_speed
227
+
228
+ pulse = generate_pulse(center_freq=frequency, n_periods=args.n_periods, sampling_freq=1.0/dt) * 1e-7
229
+
230
+ signal = np.zeros((1, int(np.ceil(np.abs(dn))) + len(pulse), 127 * 4), dtype=np.single)
231
+ for i in range(127 * 4):
232
+ signal[0, int(np.round(-dn * i / 507 if o < 0 else dn - dn * i / 507)) + np.arange(len(pulse)), i] = pulse
233
+
234
+ with h5py.File(tmp_input, 'r+') as f_i:
235
+ del f_i["u_source_index"]
236
+ f_i["u_source_index"] = (np.arange((64 + 63) * 4, dtype=np.uint64) + args.ref_offset + offset * 8).reshape(1, 1, -1).astype(np.uint64)
237
+ f_i["u_source_index"].attrs["data_type"] = np.string_(b"long")
238
+ f_i["u_source_index"].attrs["domain_type"] = np.string_(b"real")
239
+
240
+ del f_i["uy_source_input"]
241
+ f_i["uy_source_input"] = signal
242
+ f_i["uy_source_input"].attrs["data_type"] = np.string_(b"float")
243
+ f_i["uy_source_input"].attrs["domain_type"] = np.string_(b"real")
244
+
245
+ f_i["uy_source_flag"][:] = signal.shape[1]
246
+
247
+ os.system(f'../../scripts/kspaceFirstOrder-CUDA -r 20 -t 8 -g {args.device} -i {tmp_input} -o {tmp_output}')
248
+
249
+ with h5py.File(tmp_output, 'r+') as f_t:
250
+ p = np.squeeze(f_t['p'][()]).T
251
+
252
+ p = 0.35 * (p[0::4,:] + p[3::4,:]) + 0.95 * (p[1::4,:] + p[2::4,:])
253
+ T = (p.shape[1] - 1) * dt
254
+
255
+ x_orig = np.linspace(0,T,p.shape[1])
256
+ x_out = np.arange(0,T,1/40e6)
257
+
258
+ p_out = si.interp1d(x_orig, p, kind='linear', copy=False, bounds_error=False, fill_value=0, assume_sorted=True)(x_out).astype(np.float32)
259
+
260
+ with h5py.File(sim_result, 'r+') as f_o:
261
+ f_o.create_dataset(f'p_f{frequency/1e6}_o{offset}', data=p_out, compression="gzip", compression_opts=9)
262
+ except Exception as e:
263
+ print(f"Error : {e.args[0]}")
264
+ os.system(f'rm {sim_result}')
265
+
266
+ try:
267
+ os.system(f'rm {tmp_input}')
268
+ except Exception as e:
269
+ print(f"Error removing {tmp_input} : {e.args[0]}")
270
+
271
+ try:
272
+ os.system(f'rm {tmp_output}')
273
+ except Exception as e:
274
+ print(f"Error removing {tmp_output} : {e.args[0]}")
code/reference_file.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32fdd3e2950b5a13b02d6e98c2e9a55d229df362bb7a6822f9da7f0e2dcae65e
3
+ size 31711238