-
Notifications
You must be signed in to change notification settings - Fork 0
/
filter.py
496 lines (388 loc) · 20.4 KB
/
filter.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
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
import cv2
import numpy as np
import matplotlib.pyplot as plt
import plotting
import util
from scipy.spatial.distance import directed_hausdorff
import torch
import os
from sklearn.metrics.pairwise import cosine_similarity
import torch
import scipy
import scipy.fftpack
import scipy.signal
import sklearn
import knn
# two kinds of velocity?
# mean position/ time
# set distance / time
def window_average(sequence, window = 5):
average = np.convolve(sequence, np.ones(window)/window, mode='same')
return average
#input is actual coords
def search_scale_translation_multi(coords, coords_ref, save_dir = None, name = None, end = True):
window = 15
dilation = 5*15
#coords_ref_time = np.arange(coords_ref[2,:][0], coords_ref[2,:][-1])
'''
coords_ref_interpolate = torch.from_numpy(util.interpolate(coords_ref[:2,:], coords_ref[2,:], coords_ref_time))
ref_vel = torch.squeeze(util.central_diff(torch.unsqueeze(coords_ref_interpolate, dim = 1).double(), time = 1, dilation = dilation))
ref_average = window_average(np.linalg.norm(ref_vel, axis = 0), window = window)
'''
ref_dict_interp, coords_ref_time = util.interpolate_multi(coords_ref)
#print(ref_dict_interp, " multi")
ref_diff_array = util.diff_multi(ref_dict_interp, dilation = dilation)
ref_vel = util.div_multi(ref_diff_array, time = 1)
ref_average = window_average(ref_vel, window = window)
#print(ref_vel, " vel multi")
#print(ref_average, " average multi")
'''
for ra in range(100):
print(ref_dict_interp[list(ref_dict_interp.keys())[ra]], " MULTI INTERP")
'''
#for ra in range(100):
# print(ref_average[ra], " verage multi")
#print(len(ref_vel), len(ref_average), len(ref_diff_array), " ref diff array")
#print(len(list(ref_dict_interp.keys())), len(list(coords_ref_time)), " DICT LENGTH !!!!!!!!!!!!!!!!!!!")
sync_dict_interp, coords_time = util.interpolate_multi(coords)
sync_diff_array = util.diff_multi(sync_dict_interp, dilation = dilation)
#for ra in range(100):
# print(sync_dict_interp[list(sync_dict_interp.keys())[ra]], " MULTI sync INTERP")
#print(sync_dict_interp, " mutli sync_diff_array")
coords_ref_time = np.array(coords_ref_time)
coords_time = np.array(coords_time)
if save_dir is not None:
if os.path.isdir(save_dir + '/error') == False:
os.mkdir(save_dir + '/error')
os.mkdir(save_dir + '/mag_plots')
os.mkdir(save_dir + '/search')
best_scale = 1
best_shift = 0
best_correlation = -np.inf
best_sync_time = None
best_sync_average = None
#################
thresh_mag = 0.02
ref_index = np.where(ref_average > thresh_mag)[0]
#################
#print(coords_ref_time)
#print(coords_time)
#stop
'''
coords_time = np.arange(coords[2,:][0], coords[2,:][-1])
coords_interpolate = torch.from_numpy(util.interpolate(coords[:2,:], coords[2,:], coords_time))
'''
for shift in list(range(-int(coords_time.shape[0]/64), int(coords_time.shape[0]/64))):
for scale in list(np.linspace(0.5, 2, 5)) + [1.0]:
sync_time = scale*coords_time + shift
sync_vel = util.div_multi(sync_diff_array, time = scale)
#print(sync_vel, " HIII")
'''
print(np.array(sync_vel).shape, " HELLOOOO !!!!")
print(np.array(sync_time).shape, " HELLOOOO !!!!")
print(coords_time[-1], " coords time HELLOOOO !!!!")
'''
sync_average = window_average(sync_vel, window = window)
#print(np.array(sync_average).shape, " sync_averagesync_averagesync_averagesync_average !!!!")
#sync_vel = torch.squeeze(util.central_diff(torch.unsqueeze(coords_interpolate, dim = 1).double(), time = scale, dilation = dilation))
#sync_average = window_average(np.linalg.norm(sync_vel, axis = 0), window = window)
'''
for ra in range(500):
print(sync_average[ra], ra, " sync average multi")
'''
#print(sync_average, " sync average !!!!!!!!!!!!!!!!!!!!!!!!!")
sync_index = np.where(sync_average > thresh_mag)[0]
#print(sync_index, " sync index")
'''
print(ref_average.shape, " ref average")
print(coords_ref_time.shape, " coord ref trime")
print(ref_index.shape, " ref shape")
print(ref_index[-1], " last ref shape")
'''
argmins1, argmins, d2_sorted = knn.knn(np.expand_dims(coords_ref_time[ref_index], axis = 1), np.expand_dims(scale*coords_time[sync_index] + shift, axis = 1))
correlation = np.corrcoef(sync_average[sync_index][argmins], ref_average[ref_index][argmins1], rowvar = True)[0,1]
if best_correlation < correlation:
best_scale = scale
best_shift = shift
best_correlation = correlation
best_sync_time = sync_time
best_sync_average = sync_average
#print(best_sync_time)
#print(sync_average)
if save_dir is not None and end == False:
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.plot(sync_time, sync_average, c = 'r')
ax1.plot(coords_ref_time, ref_average, c = 'b')
ax1.set_title(str(correlation))
if name is not None:
fig.savefig(save_dir + '/search/' + str(name) + '_' + str(shift) + '_' + str(scale) +'_' + '.png')
else:
fig.savefig(save_dir + '/search/average_' + str(shift) + '_' + str(scale) +'_' + '.png')
plt.close('all')
'''
if save_dir is not None and end == True:
#print(len(list(ref_dict_interp.keys())), " multi coords_ref_interpolate")
#print(np.array(list(ref_dict_interp.values())).shape, "ref shape")
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.plot(list(ref_dict_interp.keys()), np.linalg.norm(np.squeeze(np.array(list(ref_dict_interp.values()))), axis = 1), c = 'r')
ax1.set_title(str("ref interpolate"))
if name is not None:
fig.savefig(save_dir + '/search/interp_' + str(name) + '_' + str(shift) + '_' + str(scale) +'_' + '.png')
else:
fig.savefig(save_dir + '/search/interp_average_' + str(shift) + '_' + str(scale) +'_' + '.png')
'''
if save_dir is not None and end == True:
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.plot(best_sync_time, best_sync_average, c = 'r')
ax1.plot(coords_ref_time, ref_average, c = 'b')
ax1.set_title(str(best_correlation))
if name is not None:
fig.savefig(save_dir + '/search/' + str(name) + '_' + str(best_shift) + '_' + str(best_scale) +'_' + '.png')
else:
fig.savefig(save_dir + '/search/average_' + str(best_shift) + '_' + str(best_scale) +'_' + '.png')
plt.close('all')
print(best_shift, best_scale, " best_shift, best_scale")
#return coords
return best_shift, best_scale, ref_dict_interp, sync_dict_interp, coords_ref_time, coords_time
#input is actual coords
def search_scale_translation(coords, coords_ref, save_dir = None, name = None, end = True):
#print(coords.shape, coords_ref.shape, " coords")
window = 15
dilation = 5*15
'''
ref_diff = np.gradient(coords_ref[:2,:], axis = 1)
sync_diff = np.gradient(coords[:2,:], axis = 1)
ref_diff_time = np.gradient(coords_ref[2,:], axis = 0)
ref_vel = np.divide(ref_diff, ref_diff_time)
'''
coords_ref_time = np.arange(coords_ref[2,:][0], coords_ref[2,:][-1])
coords_ref_interpolate = torch.from_numpy(util.interpolate(coords_ref[:2,:], coords_ref[2,:], coords_ref_time))
'''
print(coords_ref_interpolate.shape, " coords_ref_interpolate!!!!!!!!!!!!!!!!!!")
for ra in range(100):
print(coords_ref_interpolate[:, ra], " not MULTI INTERP")
'''
#print(coords_ref_interpolate, " not multi")
#print(coords_ref_interpolate.shape, " SHAPEEEE")
#print(coords_ref[:2,:].shape, " SHAEOASEEAESAES")
#ref_vel = torch.squeeze(util.central_diff(torch.unsqueeze(coords_ref[:2,:], dim = 1).double(), time = 1, dilation = dilation))
ref_vel = torch.squeeze(util.central_diff(torch.unsqueeze(coords_ref_interpolate, dim = 1).double(), time = 1, dilation = dilation))
#ref_average = (window_average(ref_vel[0, :], window = window) + window_average(ref_vel[1, :], window = window))/2.0
ref_average = window_average(np.linalg.norm(ref_vel, axis = 0), window = window)
#print(ref_vel, " not vel multi")
#for ra in range(100):
# print(ref_average[ra], " not average multi")
#print(ref_average, " not average multi")
#ref_average = ref_average[ref_ind]
if save_dir is not None:
if os.path.isdir(save_dir + '/error') == False:
os.mkdir(save_dir + '/error')
os.mkdir(save_dir + '/mag_plots')
os.mkdir(save_dir + '/search')
best_scale = 1
best_shift = 0
best_correlation = -np.inf
best_sync_time = None
best_sync_average = None
#################
'''
sync_time = coords[2,:]
sync_diff_time = np.gradient(sync_time, axis = 0)
sync_vel = np.divide(sync_diff, sync_diff_time)
#sync_average = (window_average(sync_vel[0, :], window = window) + window_average(sync_vel[1, :], window = window))/2.0
sync_average = window_average(np.linalg.norm(sync_vel, axis = 0), window = window)
sync_ind = np.where((sync_average > 0.1).all(axis=0))[0]
'''
#scale = 1.0#ref_median/sync_median
thresh_mag = 0.02#0.005
ref_index = np.where(ref_average > thresh_mag)[0]
#################
#for shift in range(0, coords_ref.shape[1] - int(coords.shape[1]/2), dilation):
coords_time = np.arange(coords[2,:][0], coords[2,:][-1])
coords_interpolate = torch.from_numpy(util.interpolate(coords[:2,:], coords[2,:], coords_time))
#for ra in range(100):
# print(coords_interpolate[:, ra], " not MULTI SYNC INTERP")
#print(coords_interpolate, " NOT mutli sync_diff_array")
for shift in list(range(-int(coords.shape[1]/64), int(coords.shape[1]/64))):
#for shift in list(range(-int(coords.shape[1]/4), int(coords.shape[1]/4))):
for scale in list(np.linspace(0.5, 2, 5)) + [1.0]:
#print(shift, scale, " H E L L O")
sync_time = scale*coords_time + shift
#sync_diff_time = np.gradient(sync_time, axis = 0)
#sync_vel = np.divide(sync_diff, sync_diff_time)*scale
print(shift, scale, " 21312qweeqdas")
sync_vel = torch.squeeze(util.central_diff(torch.unsqueeze(coords_interpolate, dim = 1).double(), time = scale, dilation = dilation))
print("********************************************************************************")
sync_average = window_average(np.linalg.norm(sync_vel, axis = 0), window = window)
for ra in range(500):
print(sync_average[ra], ra, " not sync average multi")
sync_index = np.where(sync_average > thresh_mag)[0]
'''
corr = util.find_match_time(scale*torch.from_numpy(coords_time[sync_index]) + shift, torch.from_numpy(coords_ref_time[ref_index]), 5)
argmins = list(corr[:, 0])
argmins1 = list(corr[:, 1])
'''
#print(coords_ref_time[ref_index].shape, (scale*coords_time[sync_index] + shift).shape, " HSHDAASDDS")
argmins1, argmins, d2_sorted = knn.knn(np.expand_dims(coords_ref_time[ref_index], axis = 1), np.expand_dims(scale*coords_time[sync_index] + shift, axis = 1))
correlation = np.corrcoef(sync_average[sync_index][argmins], ref_average[ref_index][argmins1], rowvar = True)[0,1]
#print(best_correlation, correlation)
if best_correlation < correlation:
best_scale = scale
best_shift = shift
best_correlation = correlation
best_sync_time = sync_time
best_sync_average = sync_average
if save_dir is not None and end == False:
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.plot(sync_time, sync_average, c = 'r')
ax1.plot(coords_ref_time, ref_average, c = 'b')
ax1.set_title(str(correlation))
if name is not None:
fig.savefig(save_dir + '/search/' + str(name) + '_' + str(shift) + '_' + str(scale) +'_' + '.png')
else:
fig.savefig(save_dir + '/search/average_' + str(shift) + '_' + str(scale) +'_' + '.png')
plt.close('all')
if save_dir is not None and end == True:
print(coords_ref_interpolate.shape, " coords_ref_interpolate")
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.plot(np.linalg.norm(coords_ref_interpolate, axis = 0), c = 'r')
ax1.set_title(str("ref interpolate"))
if name is not None:
fig.savefig(save_dir + '/search/interp_' + str(name) + '_' + str(shift) + '_' + str(scale) +'_' + '.png')
else:
fig.savefig(save_dir + '/search/interp_average_' + str(shift) + '_' + str(scale) +'_' + '.png')
if save_dir is not None and end == True:
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.plot(best_sync_time, best_sync_average, c = 'r')
ax1.plot(coords_ref_time, ref_average, c = 'b')
ax1.set_title(str(best_correlation))
if name is not None:
fig.savefig(save_dir + '/search/' + str(name) + '_' + str(best_shift) + '_' + str(best_scale) +'_' + '.png')
else:
fig.savefig(save_dir + '/search/average_' + str(best_shift) + '_' + str(best_scale) +'_' + '.png')
plt.close('all')
print(best_shift, best_scale, " best_shift, best_scale")
#return coords
return best_shift, best_scale
def average_align(coords, coords_ref, opt_iter = 5000, save_dir = None, x_lr = 1e-3, y_lr = 1e-3, t_lr = 1e-3):
"""
Iterative Closest Point
"""
coords_time = torch.tensor(list(range(coords.shape[0]))).double()
coords_ref_time = torch.tensor(list(range(coords_ref.shape[0]))).double()
rot_matrix_array = []
translation_array = []
rot_point = []
print(coords_time.shape, coords.shape, " COORDS!!!!!")
P_values = []
sync_ref = []
ref_sync = []
corr = np.correlate(coords, coords_ref,"full")
idx = np.argmax(corr);
shift = coords.shape[0] - idx
scale = np.max(coords)/np.max(coords_ref)
print(shift, scale)
coords_vec = np.stack((coords_time, coords))
coords_ref_vec = np.stack((coords_ref_time, coords_ref))
print(coords_vec.shape, coords_ref_vec.shape, " COORDS!!!!!")
corr = np.squeeze(scipy.signal.correlate2d(coords_vec, coords_ref_vec, mode='valid'))
corr_mean = np.mean(corr)
print(corr.shape, corr_mean)
stop
coords_fft = scipy.fftpack.fft(coords)
coords_ref_fft = scipy.fftpack.fft(coords_ref)
print(np.abs(coords_fft))
fig, ax1 = plt.subplots(1, 1)
ax1.plot(np.abs(coords_fft), c = 'r')
ax1.plot(np.abs(coords_ref_fft), c = 'b')
fig.savefig(save_dir + '/mag_plots/' + 'fft' + '.png')
'''
scale_x = torch.nn.Parameter(torch.tensor([1.0]).double())
scale_y = torch.nn.Parameter(torch.tensor([1.0]).double())
translate = torch.nn.Parameter(torch.tensor([0.0]).double())
#use SGD here
optimizer = torch.optim.SGD([
{'params': scale_x, 'lr': x_lr},
{'params': scale_y, 'lr': y_lr},
{'params': translate, 'lr':t_lr}
], weight_decay=0.0)
min_error = np.inf
min_params = None
error_array = []
print(coords_ref_time.shape, coords_ref.shape, " COORD REF")
coords_ref_vec = torch.unsqueeze(torch.stack((coords_ref_time, coords_ref)), dim = 0)
for t in range(opt_iter):
print(translate.repeat(coords.shape[0]).shape, " CPPPPPRDSAEW")
coords_vec = torch.unsqueeze(torch.stack((scale_x*coords_time + translate.repeat(coords.shape[0]), scale_y*coords)), dim = 0)
#loss = torch.nn.functional.conv2d(weight = coords_ref_vec)
#loss = torch.nn.CrossEntropyLoss()
print(coords_vec.shape, coords_ref_vec.shape)
error = torch.nn.functional.conv2d(input = coords_vec, weight = coords_ref_vec)#loss(coords_vec)
if min_error > error:
min_error = error.item()
if t % 1000 == 0 and save_dir is not None:
fig, ax1 = plt.subplots(1, 1)
ax1.plot(list(range(len(error_array))),error_array, c = 'blue')
if os.path.isdir(save_dir + '/error') == False:
os.mkdir(save_dir + '/error')
os.mkdir(save_dir + '/mag_plots')
#print(argmins, " argmins")
plotting.display_2d_plane([np.transpose(coords_ref.detach().cpu().numpy()), np.transpose(np.squeeze(point_transform.detach().cpu().numpy()))], save_dir + '/mag_plots', img, from_pickle, 1, 10, "iter_" + str(t) + '_' + str(i), sync = [argmins], sync1 = [argmins1])
plotting.plot_magnitude_time(np.transpose(np.array(coords_ref.detach().numpy())), np.transpose(np.squeeze(point_transform.detach().cpu().numpy())), save_dir + '/mag_plots', str(t) + '_' + str(i), sync_array = argmins, sync_array1 = argmins1)
plotting.plot_velocity(np.transpose(np.array(coords_ref.detach().numpy())), np.transpose(np.squeeze(point_transform.detach().cpu().numpy())), save_dir + '/mag_plots', str(t) + '_' + str(i), sync_array = argmins, sync_array1 = argmins1)
fig.savefig(save_dir + '/error/' + str(t) + '_.png')
#print(error)
plt.close('all')
optimizer.zero_grad()
error.backward(retain_graph=True)
optimizer.step()
#return coords
return P_values, sync_ref, ref_sync, rot_matrix_array[-1], translation_array[-1], rot_point[-1]
'''
def search_align(coords, coords_ref, save_dir = None):
"""
Iterative Closest Point
"""
coords_time = np.array(list(range(coords.shape[0])))
coords_ref_time = np.array(list(range(coords_ref.shape[0])))
rot_matrix_array = []
translation_array = []
rot_point = []
P_values = []
sync_ref = []
ref_sync = []
scale_x = 1.0
scale_y = 1.0
translate = 0.0
min_error = np.inf
min_params = None
error_array = []
print(coords_ref_time.shape, coords_ref.shape, " COORD REF")
coords_ref_vec = torch.unsqueeze(torch.stack((coords_ref_time, coords_ref)), dim = 0)
for t in range(0, coords_ref.shape[0]):
#cdist = scipy.spatial.distance.cdist( coords_time, coords_ref_time )
coords_vec = torch.unsqueeze(torch.stack((scale_x*coords_time + translate.repeat(coords.shape[0]), scale_y*coords)), dim = 0)
error = torch.nn.functional.conv2d(input = coords_vec, weight = coords_ref_vec)#loss(coords_vec)
if min_error > error:
min_error = error.item()
if t % 1000 == 0 and save_dir is not None:
fig, ax1 = plt.subplots(1, 1)
ax1.plot(list(range(len(error_array))),error_array, c = 'blue')
if os.path.isdir(save_dir + '/error') == False:
os.mkdir(save_dir + '/error')
os.mkdir(save_dir + '/mag_plots')
#print(argmins, " argmins")
plotting.display_2d_plane([np.transpose(coords_ref.detach().cpu().numpy()), np.transpose(np.squeeze(point_transform.detach().cpu().numpy()))], save_dir + '/mag_plots', img, from_pickle, 1, 10, "iter_" + str(t) + '_' + str(i), sync = [argmins], sync1 = [argmins1])
plotting.plot_magnitude_time(np.transpose(np.array(coords_ref.detach().numpy())), np.transpose(np.squeeze(point_transform.detach().cpu().numpy())), save_dir + '/mag_plots', str(t) + '_' + str(i), sync_array = argmins, sync_array1 = argmins1)
plotting.plot_velocity(np.transpose(np.array(coords_ref.detach().numpy())), np.transpose(np.squeeze(point_transform.detach().cpu().numpy())), save_dir + '/mag_plots', str(t) + '_' + str(i), sync_array = argmins, sync_array1 = argmins1)
fig.savefig(save_dir + '/error/' + str(t) + '_.png')
#print(error)
plt.close('all')
#return coords
return P_values, sync_ref, ref_sync, rot_matrix_array[-1], translation_array[-1], rot_point[-1]