-
Notifications
You must be signed in to change notification settings - Fork 8
/
corr_bench.py
447 lines (371 loc) · 18.8 KB
/
corr_bench.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
import os
import src.base_modules as pipe_base
import src.miho as miho_duplex
import src.miho_other as miho_unduplex
import src.ncc as ncc
import src.GMS.gms_custom as gms
import src.OANet.learnedmatcher_custom as oanet
import src.ACNe.acne_custom as acne
import src.AdaLAM.adalam_custom as adalam
import src.DeMatch.dematch_custom as dematch
import src.ConvMatch.convmatch_custom as convmatch
import src.DeDoDe2.dedode2_custom as dedode2
import src.FCGNN.fcgnn_custom as fcgnn
import src.CLNet.clnet_custom as clnet
import src.NCMNet.ncmnet_custom as ncmnet
import src.MS2DGNet.ms2dgnet_custom as ms2dgnet
import src.ConsensusClustering.consensusclustering_custom as consensusclustering
import src.bench_utils as bench
import numpy as np
import os
import shutil
# from src.DIM_modules.superpoint_lightglue_module import superpoint_lightglue_module
# from src.DIM_modules.disk_lightglue_module import disk_lightglue_module
# from src.DIM_modules.aliked_lightglue_module import aliked_lightglue_module
# from src.DIM_modules.loftr_module import loftr_module
def csv_write(lines, save_to='nameless.csv'):
with open(save_to, 'w') as f:
for l in lines:
f.write(l)
def compile_latex(latex_file):
# require pdflatex to be installed
os.makedirs('tmp', exist_ok=True)
shutil.copy(latex_file, 'tmp/aux.tex')
os.system('cd tmp; pdflatex aux.tex')
os.system('cd tmp; pdflatex aux.tex')
os.system('export LD_LIBRARY_PATH= && gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/printer -dNOPAUSE -dQUIET -dBATCH -dCompressFonts=true -dSubsetFonts=true -dColorConversionStrategy=/LeaveColorUnchanged -dPrinted=false -sOutputFile=tmp/aux_.pdf tmp/aux.pdf');
os.system('pdfcrop tmp/aux_.pdf tmp/aux__.pdf')
shutil.copy('tmp/aux__.pdf', latex_file[:-4] + '.pdf');
os.system('rm -R tmp');
def csv_merger(csv_list, include_match_count=False):
if not include_match_count:
avg_idx = [[ 3, 6, 'F_AUC@avg_a'], # MegaDepth
[ 6, 9, 'E_AUC@avg_a'],
[11, 14, 'F_AUC@avg_a'], # ScanNet
[14, 17, 'E_AUC@avg_a'],
[19, 22, 'H_AUC@avg_m'], # Planar
[24, 27, 'F_AUC@avg_a'], # PhotoTourism
[27, 30, 'F_AUC@avg_m'],
[30, 33, 'E_AUC@avg_a'],
[33, 36, 'E_AUC@avg_m'],
]
else:
avg_idx = [[ 4, 7, 'F_AUC@avg_a'], # MegaDepth
[ 7, 10, 'E_AUC@avg_a'],
[13, 16, 'F_AUC@avg_a'], # ScanNet
[16, 19, 'E_AUC@avg_a'],
[21, 25, 'H_AUC@avg_m'], # Planar
[28, 31, 'F_AUC@avg_a'], # PhotoTourism
[31, 34, 'F_AUC@avg_m'],
[34, 37, 'E_AUC@avg_a'],
[37, 40, 'E_AUC@avg_m'],
]
csv_data = []
for csv_file in csv_list:
aux = [csv_line.split(';') for csv_line in open(csv_file, 'r').read().splitlines()]
to_fuse = max([idx for idx, el in enumerate([s.startswith('pipe_module') for s in aux[0]]) if el == True]) + 1
tmp = {}
for row in aux:
what = ';'.join(row[:to_fuse]).replace('_outdoor_true','').replace('_outdoor_false','').replace('_fundamental_matrix','').replace('_homography','')
tmp[what] = row[to_fuse:]
csv_data.append(tmp)
pipe_set = {}
for k in csv_data:
for w in csv_data[0].keys():
pipe_set[w] = '0'
merged_csv = []
for k in pipe_set.keys():
row = [k]
for curr_csv in csv_data:
if k in curr_csv:
to_add = [el for el in curr_csv[k]]
else:
to_add = ['nan' for el in curr_csv[list(curr_csv.keys())[0]]]
row.extend(to_add)
merged_csv.append(row)
trimmed_avg_idx = []
for avg_i in avg_idx:
if avg_i[1] <= len(row):
trimmed_avg_idx.append(avg_i)
avg_csv = []
for row in merged_csv:
if 'pipe_module' in row[0]:
avg_list = [rrange[2] for rrange in trimmed_avg_idx]
else:
avg_list = [np.mean([float(i) for i in row[rrange[0]:rrange[1]]]) for rrange in trimmed_avg_idx]
avg_csv.append(avg_list)
fused_csv = []
for row_base, row_avg in zip(merged_csv, avg_csv):
row_new = []
for k in range(len(trimmed_avg_idx) - 1, - 1, - 1):
if k == 0:
l = 0
else:
l = trimmed_avg_idx[k - 1][1]
if k == len(trimmed_avg_idx) - 1:
r = len(row_base)
else:
r = trimmed_avg_idx[k][1]
row_new = row_base[l:r] + [str(row_avg.pop())] + row_new
fused_csv.append(row_new)
only_num_csv = [row[1:] for row in fused_csv[1:]]
m = np.asarray(only_num_csv, dtype=float)
sidx = np.argsort(-m, axis=0)
sidx_ = np.argsort(sidx, axis=0)
fused_csv_order = np.full((m.shape[0] + 1, m.shape[1] + 1), np.NaN)
fused_csv_order[1:,1:] = sidx_
return fused_csv, fused_csv_order
def to_latex_simple(csv_table, table_name=''):
l1 = 1
lo = 1
for row in csv_table:
l1 = max(l1, len(row[0]))
for i in row[1:]:
lo = max(lo, len(i))
header = [
'\\documentclass[a4paper,10pt]{article}\n',
'\\usepackage{graphicx}\n',
'\\usepackage{caption}\n',
'\\captionsetup{labelformat=empty}\n',
'\\usepackage{color}\n',
'\\usepackage{booktabs}\n',
'\\usepackage{amssymb}\n',
'\\usepackage[table,usenames,dvipsnames]{xcolor}\n',
'\\usepackage{amsmath}\n',
'\\usepackage{ulem}\n',
'\\usepackage{calc}\n',
'\n',
'\\newcolumntype{L}[1]{>{\\raggedright\\let\\newline\\\\\\arraybackslash\\hspace{0pt}}m{#1}}\n',
'\\newcolumntype{C}[1]{>{\\centering\\let\\newline\\\\\\arraybackslash\\hspace{0pt}}m{#1}}\n',
'\\newcolumntype{R}[1]{>{\\raggedleft\\let\\newline\\\\\\arraybackslash\\hspace{0pt}}m{#1}}\n',
'\n',
# '\\newlength\\MAXA\\setlength\\MAXA{\\widthof{' + ('a' * l1) + 'A.}}\n',
'\\newlength\\MAXA\\setlength\\MAXA{\\widthof{SuperPoint+LightGlue}}\n',
'\\newlength\\MAXB\\setlength\\MAXB{\\widthof{' + ('a' * lo) + 'A.}}\n',
'\n',
'\\begin{document}\n',
'\\pagestyle{empty}\n',
'\t\\begin{table}[t!]\n',
'\t\\centering\n',
# '\t\t\t\\begin{tabular}{R{\\MAXA}' + ('R{\\MAXB}' * (len(csv_table[0]) - 1)) + '}\n',
'\t\t\t\\begin{tabular}{R{\\MAXA}' + ('r' * (len(csv_table[0]) - 1)) + '}\n',
]
header.append('\t\t\t\t' + ' & '.join(csv_table[0]) + ' \\\\\n')
header.append('\t\t\t\t\\midrule\n')
val_table = [[int(v) for v in row[1:]] for row in csv_table[1:]]
mmax = np.argmax(np.asarray(val_table), axis=0)
mtable = [csv_table[0]]
for i, row in enumerate(csv_table[1:]):
new_row = []
for j, v in enumerate(row):
if (j != 0) and (i == mmax[j-1]):
v = '\\textbf{' + v + '}'
new_row.append(v)
mtable.append(new_row)
latex_table = []
for row in mtable[1:]:
latex_table.append('\t\t\t\t' + ' & '.join(row) + ' \\\\\n')
footer = [
'\t\t\t\end{tabular}\n',
'\t\t\\caption{' + table_name + '}\\label{none}\n',
'\t\\end{table}\n',
'\\end{document}\n',
]
latex_table = header + latex_table + footer
return latex_table
def to_latex_corr(table_name, ccol, corr_table):
header = [
'\\documentclass[a4paper,10pt]{article}\n',
'\\usepackage{graphicx}\n',
'\\usepackage{caption}\n',
'\\captionsetup{labelformat=empty}\n',
'\\usepackage{color}\n',
'\\usepackage{booktabs}\n',
'\\usepackage{amssymb}\n',
'\\usepackage[table,usenames,dvipsnames]{xcolor}\n',
'\\usepackage{amsmath}\n',
'\\usepackage{ulem}\n',
'\\usepackage{calc}\n',
'\n',
'\\newcolumntype{L}[1]{>{\\raggedright\\let\\newline\\\\\\arraybackslash\\hspace{0pt}}m{#1}}\n',
'\\newcolumntype{C}[1]{>{\\centering\\let\\newline\\\\\\arraybackslash\\hspace{0pt}}m{#1}}\n',
'\\newcolumntype{R}[1]{>{\\raggedleft\\let\\newline\\\\\\arraybackslash\\hspace{0pt}}m{#1}}\n',
'\n',
'\\newlength\\MAX\\setlength\\MAX{\\widthof{Recall.}}\n',
'\n',
'\\begin{document}\n',
'\\pagestyle{empty}\n',
'\t\\begin{table}[t!]\n',
'\t\\centering\n',
'\t\t\t\\begin{tabular}{' + ('R{\\MAX}' * (len(ccol)+1)) + '}\n',
]
header.append('\t\t\t\t' + ' & ' + ' & '.join(ccol) + ' \\\\\n')
# header.append('\t\t\t\t\\midrule\n')
latex_table = []
for i in range(len(ccol)):
row = []
for j in range(len(ccol)):
v = "{n:3.2f}".format(n=corr_table[i][j])
if i>j:
clr = 'red'
elif i < j:
clr = 'blue'
else:
clr = 'violet'
v = '\cellcolor{' + clr + '!' + str((corr_table[i][j] + 1) / 2 * 100 * 0.75 + 0.125) + '}' + v
row.append(v)
ccol[i] + ' & ' + ' & '.join(row)
latex_table.append('\t\t\t\t' + ccol[i] + ' & ' + ' & '.join(row) + ' \\\\\n')
footer = [
'\t\t\t\end{tabular}\n',
'\t\t\\caption{Error correlation for the ' + table_name + ' dataset \\textcolor{blue}{with} and \\textcolor{red}{without} MAGSAC}\\label{none}\n',
'\t\\end{table}\n',
'\\end{document}\n',
]
latex_table = header + latex_table + footer
return latex_table
if __name__ == '__main__':
pipes = [
[ 'MAGSAC^', pipe_base.magsac_module(px_th=1.00)],
[ 'MAGSACv', pipe_base.magsac_module(px_th=0.75)],
[ 'NCC', ncc.ncc_module(also_prev=True)],
[ 'MOP+MiHo', miho_duplex.miho_module()],
[ 'MOP', miho_unduplex.miho_module()],
[ 'GMS', gms.gms_module()],
[ 'OANet', oanet.oanet_module()],
[ 'AdaLAM', adalam.adalam_module()],
[ 'ACNe', acne.acne_module()],
[ 'CC', consensusclustering.consensusclustering_module()],
[ 'DeMatch', dematch.dematch_module()],
[ 'ConvMatch', convmatch.convmatch_module()],
[ 'CLNet', clnet.clnet_module()],
[ 'NCMNet', ncmnet.ncmnet_module()],
[ 'FC-GNN', fcgnn.fcgnn_module()],
['MS$^2$DG-Net', ms2dgnet.ms2dgnet_module()],
]
pipe_heads = [
[ 'SIFT+NNR', pipe_base.sift_module(num_features=8000, upright=True, th=0.95, rootsift=True)],
[ 'Key.Net+$\\scriptsize\\substack{\\text{AffNet}\\\\\\text{HardNet}}$+NNR', pipe_base.keynetaffnethardnet_module(num_features=8000, upright=True, th=0.99)],
[ 'SuperPoint+LightGlue', pipe_base.lightglue_module(num_features=8000, upright=True, what='superpoint')],
[ 'ALIKED+LightGlue', pipe_base.lightglue_module(num_features=8000, upright=True, what='aliked')],
[ 'DISK+LightGlue', pipe_base.lightglue_module(num_features=8000, upright=True, what='disk')],
[ 'LoFTR', pipe_base.loftr_module(num_features=8000, upright=True)],
[ 'DeDoDe v2', dedode2.dedode2_module(num_features=8000, upright=True)],
# [ 'SuperPoint+LightGlue (DIM)', superpoint_lightglue_module(nmax_keypoints=8000)],
# [ 'ALIKED+LightGlue (DIM)', aliked_lightglue_module(nmax_keypoints=8000)],
# [ 'DISK+LightGlue (DIM)', disk_lightglue_module(nmax_keypoints=8000)],
# [ 'LoFTR (DIM)', loftr_module(nmax_keypoints=8000)],
]
###
pipe_renamed = []
for pipe in pipes:
new_name = pipe[0]
old_name = pipe[1].get_id().replace('_outdoor_true','').replace('_outdoor_false','').replace('_fundamental_matrix','').replace('_homography','')
pipe_renamed.append([old_name, new_name])
for pipe in pipe_heads:
new_name = pipe[0]
old_name = pipe[1].get_id().replace('_outdoor_true','').replace('_outdoor_false','').replace('_fundamental_matrix','').replace('_homography','')
pipe_renamed.append([old_name, new_name])
bench_path = '../bench_data'
save_to = 'res'
benchmark_data = {
'megadepth': {'name': 'megadepth', 'Name': 'MegaDepth', 'setup': bench.megadepth_bench_setup, 'is_outdoor': True, 'is_not_planar': True, 'ext': '.png', 'use_scale': True, 'also_metric': False},
'scannet': {'name': 'scannet', 'Name': 'ScanNet', 'setup': bench.scannet_bench_setup, 'is_outdoor': False, 'is_not_planar': True, 'ext': '.png', 'use_scale': False, 'also_metric': False},
'planar': {'name': 'planar', 'Name': 'Planar', 'setup': bench.planar_bench_setup, 'is_outdoor': True, 'is_not_planar': False, 'ext': '.png', 'use_scale': False, 'also_metric': False},
'imc_phototourism': {'name': 'imc_phototourism', 'Name': 'IMC PhotoTourism', 'setup': bench.imc_phototourism_bench_setup, 'is_outdoor': True, 'is_not_planar': True, 'ext': '.jpg', 'use_scale': False, 'also_metric': True},
}
###
# # not needed if launched run_bench.py before
# for b in benchmark_data.keys():
# b_data, _ = benchmark_data[b]['setup'](bench_path=bench_path, upright=True)
csv_row_header = []
csv_col_header = None
csv_count = [[''] + [benchmark_data[b]['Name'].replace('IMC PhotoTourism','IMC-PT') for b in benchmark_data.keys()]]
all_csv = []
mask_row = []
for ip in range(len(pipe_heads)):
csv_list = []
pipe_head = pipe_heads[ip][1]
for b in benchmark_data.keys():
to_save_file = os.path.join(bench_path, save_to, save_to + '_' + pipe_head.get_id() + '_')
to_save_file_suffix ='_' + benchmark_data[b]['name']
if benchmark_data[b]['is_not_planar']:
csv_list.append(to_save_file + 'fundamental_and_essential' + to_save_file_suffix + '.csv')
else:
csv_list.append(to_save_file + 'homography' + to_save_file_suffix + '.csv')
fused_csv, _ = csv_merger(csv_list, include_match_count=True)
if csv_col_header is None:
csv_col_header = fused_csv[0]
v = pipe_head.get_id()
for renamed in pipe_renamed:
v = v.replace(renamed[0], renamed[1])
v =v.replace('_outdoor_true','').replace('_outdoor_false','')
csv_count.append([v] + [s[len('filtered_of_'):] for s in fused_csv[0] if ('filtered_of') in s])
for i in range(1, len(fused_csv)):
mask_row.append('sac' in fused_csv[i][0])
all_csv.append(fused_csv[i][1:])
csv_row_header.append(fused_csv[i][0])
num_table = np.asarray(all_csv)
mask_row = np.asarray(mask_row)
# # with no filtered field
# table_todo = [
# [ 'MegaDepth', [ 0, 1, 5, 9]],
# [ 'ScanNet', [10, 11, 15, 19]],
# [ 'PhotoTourism', [26, 27, 31, 35, 39, 43]],
# [ 'Planar', [20, 21, 25]],
# ['Non-planar', [ 0, 1, 5, 9], [10, 11, 15, 19], [26, 27, 31, 39]],
# ]
table_todo = [
[ 'MegaDepth', [ 0, 1, 2, 6, 10]],
[ 'ScanNet', [11, 12, 13, 17, 21]],
[ 'PhotoTourism', [29, 30, 31, 35, 39, 43, 47]],
[ 'Planar', [22, 23, 24, 28]],
['Non-planar', [ 0, 1, 2, 6, 10], [11, 12, 13, 17, 21], [29, 30, 31, 35, 43]],
]
for todo in table_todo:
table_name = todo[0]
idx_list = todo[1:]
num_table_fa = np.zeros((0, len(idx_list[0])))
num_table_fb = np.zeros((0, len(idx_list[0])))
for idx in idx_list:
ccol = [csv_col_header[1:][i] for i in idx]
table = num_table[:, idx]
num_table_a = num_table[mask_row][:, idx].astype(float)
num_table_a = num_table_a[np.all(np.isfinite(num_table_a), axis=1)]
num_table_fa = np.vstack((num_table_fa, num_table_a))
num_table_b = num_table[~mask_row][:, idx].astype(float)
num_table_b = num_table_b[np.all(np.isfinite(num_table_b), axis=1)]
num_table_fb = np.vstack((num_table_fb, num_table_b))
corr_table = np.triu(np.corrcoef(num_table_fa.transpose())) + np.tril(np.corrcoef(num_table_fb.transpose()), k=-1)
corr_table = np.round(corr_table * 100) / 100
for i, v in enumerate(ccol):
if 'filtered' in v: v = 'Filt.'
v = v.replace('pipeline', 'Pipeline')
v = v.replace('F_precision', 'Prec.')
v = v.replace('F_recall', 'Recall')
v = v.replace('H_precision', 'Prec.')
v = v.replace('H_recall', 'Recall')
v = v.replace('F_AUC', 'AUC$^{F}$')
v = v.replace('E_AUC', 'AUC$^{E}$')
v = v.replace('H_AUC', 'AUC$^{H}$')
v = v.replace('@5', '$_{\\text{@}5}$')
v = v.replace('@10', '$_{\\text{@}10}$')
v = v.replace('@15', '$_{\\text{@}15}$')
v = v.replace('@20', '$_{\\text{@}20}$')
v = v.replace('@(5,0.5)', '$_{\\text{@}(5,\\frac{1}{2})}$')
v = v.replace('@(10,1)', '$_{\\text{@}(10,1)}$')
v = v.replace('@(20,2)', '$_{\\text{@}(20,2)}$')
v = v.replace('@avg_a', '$_\\measuredangle$')
v = v.replace('@avg_m', '$_\\square$')
v = v.replace('$$', '')
ccol[i] = v
os.makedirs(os.path.join(bench_path, save_to, 'latex'), exist_ok=True)
latex_file = os.path.join(bench_path, save_to, 'latex', 'corr_' + table_name.lower() + '.tex')
latex_table = to_latex_corr(table_name, ccol, corr_table)
csv_write(latex_table, latex_file)
compile_latex(latex_file)
os.makedirs(os.path.join(bench_path, save_to, 'latex'), exist_ok=True)
latex_file = os.path.join(bench_path, save_to, 'latex', 'match_count.tex')
latex_table = to_latex_simple(csv_count, table_name='Average number of matches per image')
csv_write(latex_table, latex_file)
compile_latex(latex_file)