-
Notifications
You must be signed in to change notification settings - Fork 2
/
generate_result.py
768 lines (590 loc) · 27.6 KB
/
generate_result.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
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
import os
os.environ['MKL_THREADING_LAYER'] = 'GNU'
import sys
import numpy
import numpy as np
from preprocess_raw_loss_data import preprocess_file
from matplotlib import pyplot as plt
from matplotlib.ticker import LogLocator, FixedLocator, NullFormatter, FixedFormatter, NullLocator, MaxNLocator
import string
import shutil
import run_stored_cmd
ylim_thre = 0.1
success_thre = 0.2
fontsize = 20
thre = 2
cols = {'Ours': 'b',
'FD': 'g',
'SPSA': 'r',
'Ours_no_random': 'purple',
'DVG': 'orange'}
zorders = {'Ours': 1,
'FD': 2,
'SPSA': 3,
'Ours_no_random': 4,
'DVG': 5}
shader_files = [
# DVG comparisons
'apps/render_test_finite_diff_circle.py',
'apps/test_finite_diff_ring.py',
# TEG comparisons
'apps/render_test_finite_diff_circle_bw.py',
'apps/render_test_finite_diff_quadrant.py',
'apps/render_test_finite_diff_rectangle_2step.py',
# FD/SPSA comparisons
'apps/render_test_finite_diff_olympic_vec_optional_update.py',
'apps/render_test_finite_diff_ring_contour.py',
'apps/render_test_finite_diff_raymarching_siggraph_cone.py',
'apps/render_test_finite_diff_raymarching_structured_tf_logo.py',
'apps/render_test_finite_diff_raytracing_structured_tf_logo.py'
]
datas_dvg = {
'Circle':
{
'dir': 'test_finite_diff_circle',
'Ours': '_compare_diffvg',
'DVG': 'single_circle_outline_nsamples_2',
'macro': 'circle'
},
'Ring':
{
'dir': 'test_finite_diff_ring',
'Ours': '_compare_diffvg',
'DVG': 'single_circle_outline_nsamples_2',
'macro': 'circle'
}
}
datas = {
'Olympic Rings':
{
'dir': 'test_finite_diff_olympic_vec_optional_update',
'Ours': '_from_real_random',
'Ours_no_random': '_from_real_no_random',
'FD': ['_from_real_random_fd_h_%s',
'_from_real_no_random_fd_h_%s'],
'SPSA': [(1, '_from_real_random_fd_h_%s_scaled_niter'),
(65, '_from_real_random_fd_h_%s'),
(1, '_from_real_vanilla_fd_h_%s_scaled_niter'),
(40, '_from_real_vanilla_fd_h_%s'),
(1, '_from_real_no_random_fd_h_%s_scaled_niter'),
(40, '_from_real_no_random_fd_h_%s'),],
#'Ours_no_random': '_from_real5_rings',
#'FD': ['_from_real_random_5_rings_fd_h_%s'],
#'SPSA': [(1, '_from_real_random_5_rings_fd_h_%s_scaled_niter'),
# (33, '_from_real_random_5_rings_fd_h_%s'),
# (1, '_from_real_vanilla_5_rings_fd_h_%s_scaled_niter'),
# (20, '_from_real_vanilla_5_rings_fd_h_%s')],
'xmax': 52,
'same_time_SPSA': 2,
'max_halflen': 3e-3,
'ref': '/n/fs/shaderml/global_opt/proj/apps/olympic_rgb.png',
'macro': '\\olympic'
},
'Celtic Knot':
{
'dir': 'test_finite_diff_ring_contour',
'Ours': '_from_real_random',
'Ours_no_random': '_from_real_no_random',
#'FD': '_from_real_random_fd_%s',
'FD': ['_from_real_random_fd_%s',
'_from_real_no_random_fd_%s'],
'SPSA': [(1, '_from_real_random_fd_%s_scaled_niter'),
(42, '_from_real_random_fd_%s'),
(1, '_from_real_vanilla_fd_h_%s_scaled_niter'),
(21, '_from_real_vanilla_fd_h_%s'),
(1, '_from_real_no_random_fd_%s_scaled_niter'),
(21, '_from_real_no_random_fd_%s')],
'xmax': 7,
'same_time_SPSA': 2,
'max_halflen': 2e-3,
'ref': '/n/fs/shaderml/differentiable_compiler/celtic_knot.png',
'macro': '\\celtic'
},
'SIGGRAPH':
{
'dir': 'test_finite_diff_raymarching_siggraph_cone',
'Ours': '_from_real_random',
'Ours_no_random': '_from_real_no_random',
#'FD': '_from_real_random_fd_h_%s',
'FD': ['_from_real_random_fd_h_%s',
'_from_real_no_random_fd_h_%s'],
'SPSA': [(1, '_from_real_random_fd_h_%s_scaled_niter'),
(27, '_from_real_random_fd_h_%s'),
(1, '_from_real_vanilla_fd_h_%s_scaled_niter'),
(22, '_from_real_vanilla_fd_h_%s'),
(1, '_from_real_no_random_fd_h_%s_scaled_niter'),
(22, '_from_real_no_random_fd_h_%s')],
'xmax': 70,
'same_time_SPSA': 2,
'max_halflen': 2e-3,
'ref': '/n/fs/shaderml/global_opt/proj/apps/siggraph_gradient.png',
'macro': '\\siggraph'
},
'TF Raymarch':
{
'dir': 'test_finite_diff_raymarching_structured_tf_logo_updated',
'Ours': '_from_real',
'FD': ['_from_real_fd_h_%s'],
'SPSA': [(1, '_from_real_fd_h_%s_scaled_niter'),
(16, '_from_real_fd_h_%s'),
(1, '_from_real_vanilla_fd_h_%s_scaled_niter'),
(16, '_from_real_vanilla_fd_h_%s')],
'xmax': 10,
'same_time_SPSA': 1,
'max_halflen': 2e-3,
'ref': '/n/fs/shaderml/differentiable_compiler/tf_logo.png',
'macro': '\\tfmarch'
},
'TF Raycast':
{
'dir': 'test_finite_diff_raytracing_structured_tf_logo',
'Ours': '_from_real_random',
'Ours_no_random': '_from_real_no_random',
#'FD': '_from_real_random_fd_%s',
#'FD': '_from_real_vanilla_fd_h_%s',
'FD': ['_from_real_random_fd_%s',
'_from_real_no_random_fd_%s',
'_from_real_vanilla_fd_h_%s'],
'SPSA': [(1, '_from_real_random_fd_%s_scaled_niter'),
(27, '_from_real_random_fd_%s'),
(1, '_from_real_vanilla_fd_h_%s_scaled_niter'),
(16, '_from_real_vanilla_fd_h_%s'),
(1, '_from_real_no_random_fd_%s_scaled_niter'),
(16, '_from_real_no_random_fd_%s')],
'xmax': 20,
'same_time_SPSA': 4,
'max_halflen': 2e-3,
'ref': '/n/fs/shaderml/differentiable_compiler/tf_logo.png',
'macro': '\\tfcast'
},
}
table_header = """
\\begin{tabular}{c|c|c|c|c|c|c|c|c}
\\multirow{2}{*}{Shader} & \\multicolumn{4}{c|}{Med. Success Time} & \\multicolumn{4}{c}{Exp. Time to Success} \\\\ \\cline{2-9}
& Ours & \\Owo & FD$^*$ & SPSA$^*$ & Ours & \\Owo & FD$^*$ & SPSA$^*$ \\\\ \\thickhline
"""
table_end = """
\\end{tabular}
"""
metric_table_header = """
\\begin{tabular}{cccc}
& Ours & FD & SPSA \\\\
"""
result_table_header = """
\\begin{tabular}{ccc}
Target & Optimization & Modified \\\\
"""
result_table_tf_header = """
\\begin{tabular}{ccc}
"""
def scientific_format(val):
base_strs = ('%.2e' % val).split('e')
ans = '$%s \\times 10^{%d}$' % (base_strs[0], int(base_strs[1]))
return ans
def get_median(loss):
nvalid = loss.shape[1] - np.isnan(loss).sum(-1)
max_nvalid = np.max(nvalid)
last_loss = loss[np.arange(loss.shape[0]), nvalid - 1]
need_fill_mask = np.isnan(loss[:, :max_nvalid])
values_to_fill = np.tile(np.expand_dims(last_loss, 1), [1, max_nvalid])
loss[:, :max_nvalid][need_fill_mask] = values_to_fill[need_fill_mask]
median_val = np.median(loss, 0, keepdims=True)
return median_val
def process_log(file):
lines = open(file).read().split('\n')
success = False
found = False
for line in lines[::-1]:
if line.startswith('runtime per iter'):
runtime = -float(line.replace('runtime per iter', ''))
success = False
found = True
if line.startswith('99 '):
if found:
success = True
break
else:
try:
val = float(line[:2])
success = False
found = False
except:
pass
assert success
assert found
return runtime
def draw_subplot(ax, shader, count, err_datas, success_median_times=None):
#dummy = success_median_times
#success_median_times = None
L_min = 1e8
L_max = -1e8
for method in err_datas.keys():
loss, *_ = err_datas[method]
L_min = min(L_min, np.nanmin(loss))
L_max = max(L_max, np.nanmax(loss))
current_thre = thre
if success_median_times is not None:
L_min_ours = np.nanmin(err_datas['Ours'][0])
current_thre = thre * L_min_ours / L_min
ax.plot(np.arange(datas[shader]['xmax'] + 1), current_thre * np.ones_like(np.arange(datas[shader]['xmax'] + 1)), 'gray', linewidth=2)
#L_min = np.nanmin(err_datas['Ours'][0])
k = (L_max / L_min) ** (1 / 3)
for method in err_datas.keys():
if method in ['Ours', 'DVG', 'Ours_no_random']:
label = method.replace('_', ' ')
else:
label = '%s$^*$' % method
loss, scale, *_ = err_datas[method]
loss /= L_min
iterations = np.arange(loss.shape[1]) + 1
for idx in range(loss.shape[0]):
ax.plot(iterations * scale, loss[idx], cols[method], alpha=0.05, zorder=zorders[method])
median_val = get_median(loss)[0]
ax.plot(iterations * scale, median_val, 'w', alpha=0.5, linewidth=5, zorder=zorders[method])
ax.plot(iterations * scale, median_val, cols[method], label=label, zorder=zorders[method])
if success_median_times is not None:
for method in list(err_datas.keys())[::-1]:
success_median_time = err_datas[method][3]
ax.scatter(success_median_time, current_thre, c=cols[method], s=50, zorder=200+zorders[method])
#for idx in range(len(success_median_times)):
# ax.scatter(success_median_times[idx], thre, c=list(cols.values())[idx], zorder=200, s=50)
ax.set_yscale('log')
# Disable minor ticks, because they will be inconsistent for different k values
#nminor_ticks = 10
#ax.yaxis.set_minor_locator(LogLocator(base=k, subs=[i * k / nminor_ticks for i in range(1, nminor_ticks)]))
#ax.yaxis.set_minor_formatter(NullFormatter())
#ax.yaxis.set_minor_locator(NullLocator())
if False:
yticks = k ** np.arange(4)
ax.yaxis.set_major_locator(FixedLocator(yticks))
if count == 0:
ax.yaxis.set_major_formatter(FixedFormatter(['$\epsilon^%d$' % idx for idx in range(4)]))
ax.legend(fontsize=fontsize)
else:
ax.yaxis.set_major_formatter(NullFormatter())
else:
yticks = 10 ** np.arange(np.ceil(np.log10(L_max / L_min)))
ax.yaxis.set_major_locator(FixedLocator(yticks))
ax.yaxis.set_major_formatter(FixedFormatter(['%d' % idx for idx in range(yticks.shape[0])]))
if shader in ['SIGGRAPH', 'Circle']:
ax.legend(fontsize=fontsize)
if success_median_times:
if count in [1, 3]:
ax.xaxis.set_major_locator(MaxNLocator(nbins=3, integer=True))
if count == 0:
ax.set_ylabel('Error (order of magnitude)', fontsize=fontsize)
#ax.yaxis.set_label_position("right")
else:
ax.xaxis.set_major_locator(MaxNLocator(nbins=4, integer=True))
if count == 0:
ax.set_ylabel('Error (order of magnitude)', fontsize=fontsize)
#ax.yaxis.set_label_position("right")
if success_median_times is not None:
# force last tick to dissapear
ax.set_xlim(0, datas[shader]['xmax'] - 1e-8)
ax.set_ylim(1 / k ** ylim_thre, L_max / L_min * k ** ylim_thre)
if not shader.startswith('TF'):
titlename = shader
else:
if shader == 'TF Raymarch':
titlename = 'TF RayMarch'
else:
titlename = 'TF RayCast'
if success_median_times is None:
pass
#ax.set_title('(' + string.ascii_lowercase[count] + ') ' + titlename, fontsize=fontsize, y=-0.2)
#ax.set_title('(' + string.ascii_lowercase[count] + ') ' + titlename, fontsize=fontsize, y=-0.2)
ax.tick_params(axis='both', which='major', labelsize=fontsize)
def generate_result():
parent_dir = sys.argv[1]
save_dir = os.path.join(parent_dir, 'result')
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
fig, axes = plt.subplots(nrows=1, ncols=len(datas_dvg), figsize=(5 * len(datas_dvg), 5))
count = 0
for shader in datas_dvg.keys():
current_dir = os.path.join(parent_dir, datas_dvg[shader]['dir'])
err_datas = {}
for method in ['Ours', 'DVG']:
if method == 'Ours':
file = 'ours_both_sides_1_scale_L2_adam_1.0e-02%s_all_loss.npy' % datas_dvg[shader][method]
processed_file = os.path.join(current_dir, 'processed_' + file)
file = os.path.join(current_dir, file)
logfile = os.path.join(current_dir, 'log_ours_both_sides_1_scale_L2_adam_1.0e-02%s.txt' % datas_dvg[shader][method])
scale = process_log(logfile)
else:
file = os.path.join(current_dir, '../diffvg/results/single_circle_nsamples_2/all_loss.npy')
processed_file = os.path.join(current_dir, 'processed_all_loss.npy')
logfile = os.path.join(current_dir, '../diffvg/results/single_circle_nsamples_2/log.txt')
scale = -process_log(logfile)
if os.path.exists(processed_file):
loss = np.load(processed_file)
else:
loss = preprocess_file(file)
np.save(processed_file, loss)
err_datas[method] = (loss,
scale,
file.replace('_all_loss.npy', ''))
ax = axes[count]
draw_subplot(ax, shader, count, err_datas)
count += 1
plt.subplots_adjust(wspace=0.1)
plt.savefig(os.path.join(save_dir, 'error_plot_DVG.png'), bbox_inches='tight')
plt.close()
fig, axes = plt.subplots(nrows=1, ncols=len(datas), figsize=(5 * len(datas), 5))
if len(datas) == 1:
axes = np.array([axes])
count = 0
table_str = table_header
for shader in datas.keys():
current_dir = os.path.join(parent_dir, datas[shader]['dir'])
err_datas = {}
L_min = 1e8
L_max = -1e8
for method in ['Ours', 'Ours_no_random', 'FD', 'SPSA']:
loss_files = []
log_files = []
if method not in datas[shader].keys():
continue
if method.startswith('Ours'):
loss_files.append('ours_both_sides_5_scale_L2_adam_1.0e-02%s_all_loss.npy' % datas[shader][method])
log_files.append('log_ours_both_sides_5_scale_L2_adam_1.0e-02%s.txt' % datas[shader][method])
elif method == 'FD':
for suffix in datas[shader][method]:
for h_exp in range(5):
h_str = '0' * (h_exp + 1) + '1'
loss_files.append('finite_diff_both_sides_5_scale_L2_adam_1.0e-02%s_all_loss.npy' % (suffix % h_str))
log_files.append('log_finite_diff_both_sides_5_scale_L2_adam_1.0e-02%s.txt' % (suffix % h_str))
else:
for nsamples, suffix in datas[shader][method]:
for h_exp in range(5):
h_str = '0' * (h_exp + 1) + '1'
loss_files.append('finite_diff%d_both_sides_5_scale_L2_adam_1.0e-02%s_all_loss.npy' % (nsamples, suffix % h_str))
log_files.append('log_finite_diff%d_both_sides_5_scale_L2_adam_1.0e-02%s.txt' % (nsamples, suffix % h_str))
current_min = 1e8
current_loss = None
current_log = None
for idx in range(len(loss_files)):
file = loss_files[idx]
processed_file = os.path.join(current_dir, 'processed_' + file)
if os.path.exists(processed_file):
loss = np.load(processed_file)
else:
loss = preprocess_file(os.path.join(current_dir, file))
np.save(processed_file, loss)
median_val = get_median(loss)
#min_val = np.nanmin(median_val)
min_val = np.nanmin(loss)
if min_val < current_min:
current_loss = loss
current_log = log_files[idx]
current_min = min_val
if np.nanmin(loss) < L_min and method != 'Ours_no_random':
L_min = np.nanmin(loss)
if np.nanmax(loss) > L_max:
L_max = np.nanmax(loss)
err_datas[method] = (current_loss,
process_log(os.path.join(current_dir, current_log)),
file.replace('_all_loss.npy', ''))
datas[shader]['err'] = err_datas
# generate convergence plot and table
#thre = (L_max / L_min) ** success_thre
ax = axes[count]
success_rates = []
success_median_times = []
expected_times_to_success = []
expected_nsamples = 100
for method in err_datas.keys():
if method.startswith('Ours'):
label = method.replace('_', ' ')
else:
label = '%s$^*$' % method
loss, scale, _ = err_datas[method]
loss /= L_min
iterations = np.arange(loss.shape[1]) + 1
success_rate = np.sum(np.nanmin(loss, 1) < thre)
success_min_idx = np.argmax(loss < thre, 1)
if np.sum(success_min_idx[success_min_idx != 0]):
median_idx = np.median(success_min_idx[success_min_idx != 0])
else:
median_idx = np.inf
success_median_time = median_idx * scale
success_rates.append(success_rate)
success_median_times.append(success_median_time)
if success_rate > 0:
sum_time = 0
for _ in range(expected_nsamples):
err = 1e8
while err > thre:
current_run = loss[np.random.choice(np.arange(loss.shape[0]))]
if np.nanmin(current_run) < thre:
sum_time += np.where(current_run < thre)[0][0]
break
else:
sum_time += np.sum(np.logical_not(np.isnan(current_run)))
avg_time = sum_time * scale / expected_nsamples
else:
avg_time = np.inf
print(method, success_rate, success_median_time, avg_time)
expected_times_to_success.append(avg_time)
err_datas[method] += (success_median_time, avg_time)
metric_median = []
metric_expected = []
min_idx_median = None
min_idx_expected = None
min_median = 1e8
min_expected = 1e8
method_count = 0
for method in err_datas.keys():
metric_median.append('%.1f'% err_datas[method][3] if not np.isinf(err_datas[method][3]) else '\\na')
metric_expected.append('%.1f'% err_datas[method][4] if not np.isinf(err_datas[method][4]) else '\\na')
if err_datas[method][3] < min_median:
min_median = err_datas[method][3]
min_idx_median = method_count
if err_datas[method][4] < min_expected:
min_expected = err_datas[method][4]
min_idx_expected = method_count
method_count += 1
metric_median[min_idx_median] = '\\textbf{%s}' % metric_median[min_idx_median]
metric_expected[min_idx_expected] = '\\textbf{%s}' % metric_expected[min_idx_expected]
if shader == 'TF Raymarch':
# Ours without random should be identical with Ours with random
metric_median = metric_median[:1] * 2 + metric_median[1:]
metric_expected = metric_expected[:1] * 2 + metric_expected[1:]
time_str = ' & '.join(metric_median)
expected_time_str = ' & '.join(metric_expected)
#time_str = '\\textbf{%.1f} &' % err_datas['Ours'][3] + ' & '.join(['%.1f'% val if not np.isinf(val) else '\\na' for val in success_median_times[1:2]])
#expected_time_str = '\\textbf{%.1f} &' % expected_times_to_success[0] + ' & '.join(['%.1f' % val if not np.isinf(val) else '\\infin' for val in expected_times_to_success[1:2]])
shadername = datas[shader]['macro']
table_str += f"""
{shadername} & {time_str} & {expected_time_str} \\\\ \\hline
"""
#if 'Ours_no_random' in err_datas.keys():
# del err_datas['Ours_no_random']
draw_subplot(ax, shader, count, err_datas, success_median_times)
count += 1
plt.subplots_adjust(wspace=0.1)
plt.savefig(os.path.join(save_dir, 'error_plot.png'), bbox_inches='tight')
plt.close()
table_str += table_end
open(os.path.join(save_dir, 'err_table.txt'), 'w').write(table_str)
# generate error metric figures and table
metric_table_str = metric_table_header
for shader in datas.keys():
current_dir = os.path.join(parent_dir, datas[shader]['dir'])
rhs = np.load(os.path.join(current_dir,
'random_smooth_metric_2X100000_len_%f_2D_kernel_rhs.npy' % datas[shader]['max_halflen']))
if shader == 'Olympic Rings':
metric_table_str += """\\multirow{2}{*}{\\raisebox{1.8\\normalbaselineskip}[0pt][0pt]{\\rotatebox[origin=c]{90}{%s}}} """ % '\\olympicshort'
else:
metric_table_str += """\\raisebox{2.5\\normalbaselineskip}[0pt][0pt]{\\rotatebox[origin=c]{90}{%s}}""" % datas[shader]['macro']
min_errs = []
for method in ['Ours', 'FD', 'SPSA']:
if method == 'Ours':
suffixes = ['_2D_kernel']
elif method == 'FD':
suffixes = ['_2D_kernel_FD_finite_diff_%f' % 0.1 ** h for h in range(1, 6)]
else:
suffixes = ['_2D_kernel_SPSA_' + str(datas[shader]['same_time_SPSA']) + '_finite_diff_%f' % 0.1 ** h for h in range(1, 6)]
min_err = 1e8
min_map = None
for suffix in suffixes:
lhs = np.load(os.path.join(
current_dir,
'kernel_smooth_metric_debug_10000X1_len_%f_kernel_box_sigma_1.000000_0.100000%s.npy' %
(datas[shader]['max_halflen'], suffix)))
err_map = np.abs(lhs - rhs).transpose()
#err = np.mean(err_map[err_map != 0])
err = np.mean(err_map)
if err < min_err:
min_err = err
min_map = err_map
if shader == 'Olympic Rings':
pad = 20
min_map2 = np.zeros((min_map.shape[0] + 2 * pad, min_map.shape[1] + 2 * pad))
min_map2[pad:-pad, pad:-pad] = min_map
min_map = min_map2
min_err = np.mean(min_map)
min_errs.append(min_err)
plt.figure()
plt.imshow(min_map, vmin=0, vmax=0.01, cmap='hot')
plt.axis('off')
err_name = '%s_%s.png' % (shader, method)
plt.savefig(os.path.join(save_dir, err_name), bbox_inches='tight')
plt.close()
metric_table_str += """& \\includegraphics[width=\\w]{figures/metric/%s} """ % err_name
#err_str = ' & '.join([''] + ['%.2e' % err for err in min_errs])
#err_str = ' & '.join([''] + [scientific_format(err) for err in min_errs])
err_str = ' & '.join([''] + [scientific_format(min_errs[idx]) if idx == 0 else '%.1fx' % (min_errs[idx] / min_errs[0]) for idx in range(len(min_errs))])
metric_table_str += """ \\\\
%s \\\\
""" % err_str
metric_table_str += """& \\multicolumn{3}{c}{
\\includegraphics[width=3\\w]{figures/metric/colorbar.png}}
""" + table_end
open(os.path.join(save_dir, 'metric_table.txt'), 'w').write(metric_table_str)
dummy = np.array([[0,1]])
plt.figure(figsize=(9, 0.5))
img = plt.imshow(dummy, cmap="hot", vmin=0, vmax=0.01)
plt.gca().set_visible(False)
cax = plt.axes([0.1, 0.2, 0.8, 0.6])
cbar = plt.colorbar(orientation="horizontal", cax=cax)
cbar.set_ticks([0, 0.01])
cbar.ax.tick_params(labelsize=fontsize)
plt.savefig(os.path.join(save_dir, "colorbar.png"), bbox_inches='tight')
# generate result figures and table
result_table = result_table_header
result_table_tf_raymarch = result_table_tf_header
result_table_tf_raytrace = result_table_tf_header
for shader in datas.keys():
current_dir = os.path.join(parent_dir, datas[shader]['dir'])
loss, scale, prefix, *_ = datas[shader]['err']['Ours']
min_loss = np.nanmin(loss, 1)
# first 5 might be modified by other processes
sorted_idx = np.argsort(min_loss[5:]) + 5
opt_file = os.path.join(current_dir, prefix + '_result%d_0.png' % sorted_idx[0])
opt_dst = os.path.join(save_dir, 'opt_%s.png' % shader)
if not os.path.exists(opt_dst):
shutil.copyfile(opt_file, opt_dst)
ref_dst = os.path.join(save_dir, 'ref_%s.png' % shader)
if not os.path.exists(ref_dst):
shutil.copyfile(datas[shader]['ref'], ref_dst)
fig_str = """
\\includegraphics[width=\\w]{figures/result/ref_%s.png} & \\includegraphics[width=\\w]{figures/result/opt_%s.png} & \\includegraphics[width=\\w]{figures/result/anim_%s.png} \\\\
""" % (shader, shader, shader)
if shader == 'TF Raycast':
result_table_tf_raytrace += fig_str + """
(a) Opt & (b) Modified & (c) Modified \\\\
"""
elif shader == 'SIGGRAPH':
# pass because already include its result in teaser figure
pass
else:
result_table += fig_str + """
%s & & \\\\
""" % datas[shader]['macro']
result_table_tf_raymarch += table_end
result_table_tf_raytrace += table_end
result_table += table_end
#open(os.path.join(save_dir, 'result_table_tf_raymarch.txt'), 'w').write(result_table_tf_raymarch)
open(os.path.join(save_dir, 'result_table_tf_raytrace.txt'), 'w').write(result_table_tf_raytrace)
open(os.path.join(save_dir, 'result_table.txt'), 'w').write(result_table)
def collect_result():
for file in shader_files:
run_stored_cmd.run(file, sys.argv[1])
def main():
error = False
if len(sys.argv) != 3:
error = True
elif sys.argv[2] not in ['generate', 'collect']:
error = True
if error:
print('Usage: python generate_result <path> <mode>')
print('mode = {generate, collect}')
return
if sys.argv[2] == 'generate':
generate_result()
else:
collect_result()
if __name__ == '__main__':
main()