forked from jswulff/mrflow
-
Notifications
You must be signed in to change notification settings - Fork 0
/
mrflow.py
385 lines (294 loc) · 13.2 KB
/
mrflow.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
#! /usr/bin/env python2
import sys,os
import argparse
import time
import numpy as np
import cv2
MRFLOW_HOME = os.environ['MRFLOW_HOME']
sys.path.append(MRFLOW_HOME)
from utils import print_exception as pe
from utils import compute_figure
from utils import flow_io as fio
from utils import flow_viz as fviz
from utils import structure2image
import load_data
import parameters
from pipeline import match_features
from pipeline import initial_alignment
from pipeline import compute_structure
from pipeline import optimize_variational_refinement as optimize
from pipeline import structure2flow
from pipeline import evaluate_flow
def compute_mrflow(paths,params):
""" Process a single frame.
"""
tdir = params.tempdir
print('(MM) PATHS:')
for k in sorted(paths.keys()):
print('(MM)\t {}: {}'.format(k,paths[k]))
# Load data
data_in = load_data.load_initial_data(params=params, **paths)
override = False
if 'error_override' in data_in.keys() and data_in['error_override'] is True:
print('(EE) Overriding computation...')
u,v = data_in['flow'][1]
flow = data_in['flow']
override=True
rigidity_output = np.ones_like(u)>0
structure_optimized = np.ones_like(u)
else:
try:
# Load data from input structure
flow = data_in['flow']
#flow_fwd_gt = data_in['flow_fwd_gt'] if 'flow_fwd_gt' in data_in.keys() else None
rigidity_thresholded = data_in['rigidity_thresholded']
rigidity = data_in['rigidity']
images = data_in['images']
occlusions = data_in['occlusions']
# params = data_in['params'].tolist()
except:
print('Exception occured in file {}'.format(fil))
pe.print_exception()
raise Exception()
if params.debug_use_rigidity_gt:
rigidity_thresholded = data_in['rigidity_gt'] > 0.5
rigidity = data_in['rigidity_gt'] > 0.5
try:
# Step 2: Feature matching
tmatch_0 = time.time()
(matches_pairwise,
matches_all_frames,
outliers_large_pairwise,
outliers_F_pairwise) = match_features.generate_features(
images, flow, occlusions, rigidity_thresholded, params)
tmatch_1 = time.time()
tinit_0 = time.time()
# Step 3: Alignment
(homographies,
epipoles,
plane_global) = initial_alignment.align_frames(
images,
matches_all_frames,
matches_pairwise,
params)
# Step 4: Building structure
(structure,
homographies_refined,
mu,
B,
epipoles_new,
rigidity_refined,
occlusions) = compute_structure.compute_structure(
images,
flow,
rigidity,
occlusions,
homographies,
epipoles,
params)
tinit_1 = time.time()
topt_0 = time.time()
# Step 5: Optimize structure
(structure_optimized,) = optimize.optimize(
images,
structure,
rigidity_refined,
epipoles_new,
homographies_refined,
B,
mu,
occlusions,
params)
topt_1 = time.time()
tcomb_0 = time.time()
# Step 6: Re-compute flow.
rigidity_output, u,v = structure2flow.combine_flow(
structure_optimized, # A
flow[1], # flow_init
rigidity_refined,
epipoles_new,
mu,
homographies_refined,
B,
images,
params)
tcomb_1 = time.time()
print('=============== Timing information ====================')
print('Feature matching: \t {0:2.3f} seconds'.format(tmatch_1-tmatch_0))
print('Alignment & rigidity computation: \t {0:2.3f} seconds'.format(tinit_1-tinit_0))
print('Optimization: \t {0:2.3f} seconds'.format(topt_1-topt_0))
print('Merging: \t {0:2.3f} seconds'.format(tcomb_1-tcomb_0))
print('=======================================================')
except:
# Some mistake occured.
print('==== EXCEPTION ====')
pe.print_exception()
# Just use initial flow.
u,v = flow[1]
rigidity_output = np.ones_like(u)>0
structure_optimized = np.ones_like(u)
#
# Write flow and images
#
# Save flow output
fio.flow_write(params.tempdir + '/' + 'flow.flo', u, v)
# Save structure output
np.save(params.tempdir + '/' + 'structure.npy', structure_optimized)
# fio.flow_write_png(u, v, params.tempdir + '/' + 'flow.png')
Iest = fviz.computeFlowImage(u,v)
uinit,vinit = flow[1]
Iinit = fviz.computeFlowImage(uinit,vinit)
if override:
rigidity_output = np.ones_like(u)>0
rigidity_viz = (np.dstack((rigidity_output,rigidity_output,rigidity_output))*255).astype('uint8')
structure_viz = structure2image.structure2image(structure_optimized,
rigidity_output,
cmap='viridis')
flowdiff = np.sqrt((u-uinit)**2 + (v-vinit)**2)
flowdiff = 255.0 * flowdiff / max(1.0,flowdiff.max())
flowdiff = np.dstack((flowdiff, flowdiff, flowdiff)).astype('uint8')
Iviz = np.vstack(( np.hstack((Iinit, Iest)), np.hstack((flowdiff, rigidity_viz)) ))
# Save visualization image
def saveim(fname, I):
if I.ndim == 3:
cv2.imwrite(fname, I[:,:,::-1])
else:
cv2.imwrite(fname, I)
saveim(params.tempdir + '/comparison.png', Iviz)
saveim(params.tempdir + '/structure.png', structure_viz)
saveim(params.tempdir + '/flow_viz.png', Iest)
# Save rigidity output
saveim(params.tempdir + '/rigidity.png', (rigidity_output>0).astype('uint8')*255)
## Compute errors
if 'flow_fwd_gt' in data_in and data_in['flow_fwd_gt'] is not None:
ugt,vgt = data_in['flow_fwd_gt']
flow_gt_valid = data_in['flow_fwd_gt_valid']
uinit,vinit = data_in['flow'][1]
# Do we have a GT rigidity map?
if 'rigidity_gt' in data_in.keys() and data_in['rigidity_gt'].tolist() is not None:
rigidity_gt = data_in['rigidity_gt']
else:
rigidity_gt = np.ones_like(ugt)>0
if 'flow_fwd_gt_valid' in data_in.keys() and data_in['flow_fwd_gt_valid'].tolist() is not None:
flow_fwd_gt_valid = data_in['flow_fwd_gt_valid']
else:
flow_fwd_gt_valid = np.ones_like(ugt)>0
# Compute error of estimated flow
(epe_est_rigid,
epe_est_nonrigid,
epe_est_all,
perc_est_rigid,
perc_est_nonrigid,
perc_est_all) = evaluate_flow.compute_errors(ugt,vgt,u,v,rigidity_gt,valid=flow_fwd_gt_valid)
# Compute error of initial flow
(epe_init_rigid,
epe_init_nonrigid,
epe_init_all,
perc_init_rigid,
perc_init_nonrigid,
perc_init_all) = evaluate_flow.compute_errors(ugt,vgt,flow[1][0],flow[1][1],rigidity_gt,valid=flow_fwd_gt_valid)
#
# Display errors
#
print('******************** EPE STATS ********************')
print('\t\t\tSTATIC \tNON-STATIC\tALL')
print('Initial flow: \t{0:2.5f} \t{1:2.5f} \t{2:2.5f}'.format(epe_init_rigid,epe_init_nonrigid,epe_init_all))
print('Estimated flow: \t{0:2.5f} \t{1:2.5f} \t{2:2.5f}'.format(epe_est_rigid,epe_est_nonrigid,epe_est_all))
print('***************************************************')
print('******************** PERC STATS ********************')
print('\t\t\tSTATIC \tNON-STATIC\tALL')
print('Initial flow: \t{0:2.5f} \t{1:2.5f} \t{2:2.5f}'.format(perc_init_rigid,perc_init_nonrigid,perc_init_all))
print('Estimated flow: \t{0:2.5f} \t{1:2.5f} \t{2:2.5f}'.format(perc_est_rigid,perc_est_nonrigid,perc_est_all))
print('***************************************************')
if vars(params).has_key('debug_compute_figure') and vars(params)['debug_compute_figure'] == 1:
# Compute teaser image
compute_figure.plot_figure_1(images, rigidity_refined, structure_optimized, (u,v), (ugt,vgt))
if vars(params).has_key('debug_compute_figure') and vars(params)['debug_compute_figure'] == 2:
compute_figure.plot_figure_2(images, flow, rigidity, structure[1], occlusions, rigidity_refined, structure_optimized, (u,v))
if vars(params).has_key('debug_compute_figure') and vars(params)['debug_compute_figure'] == 5:
compute_figure.plot_figure_5(images, rigidity_refined, structure_optimized, (u,v), (uinit,vinit), (ugt,vgt), flow_gt_valid )
if vars(params).has_key('debug_compute_figure') and vars(params)['debug_compute_figure'] == 6:
# Compute supmat results figure
compute_figure.plot_figure_6(images, rigidity_refined, structure_optimized, (u,v), (uinit,vinit), (ugt,vgt), flow_gt_valid)
#
# Output error images
#
if params.debug_save_frames:
evaluate_flow.save_frames(
u,
v,
flow[1][0],
flow[1][1],
ugt,
vgt,
flow_fwd_gt_valid>0,
rigidity_output,
params)
#
# Output CSV file
#
data_array = [ (epe_init_rigid, epe_est_rigid),
(epe_init_nonrigid, epe_est_nonrigid),
(epe_init_all, epe_est_all),
(perc_init_rigid, perc_est_rigid),
(perc_init_nonrigid, perc_est_nonrigid),
(perc_init_all, perc_est_all)]
names = ['epe_rigid.csv',
'epe_nonrigid.csv',
'epe_all.csv',
'perc_rigid.csv',
'perc_nonrigid.csv',
'perc_all.csv']
for csv_data, csv_name in zip(data_array, names):
fname = os.path.join(params.tempdir, csv_name)
with open(fname, 'w') as fil:
fil.write(',InitialFlow,MR-Flow\n')
fil.write('{},{},{}\n'.format(params.tempdir, csv_data[0],csv_data[1]))
def main():
parser = argparse.ArgumentParser()
# Required arguments
parser.add_argument('frame0', type=str, help='Frame at T-1')
parser.add_argument('frame1', type=str, help='Frame at T')
parser.add_argument('frame2', type=str, help='Frame at T+1')
# Optional arguments
parser.add_argument('--rigidity', type=str, default='', help='Rigidity initialization')
parser.add_argument('--flow_fwd', type=str, default='', help='Initial flow in forward direction (T to T+1)')
parser.add_argument('--backflow_fwd', type=str, default='', help='Initial backflow in forward direction (T+1 to T)')
parser.add_argument('--flow_bwd', type=str, default='', help='Initial flow in backward direction (T to T-1)')
parser.add_argument('--backflow_bwd', type=str, default='', help='Initial backflow in backward direction (T-1 to T)')
# Override initialization arguments
parser.add_argument('--no_init', action='store_true', help='Do not provide initial rigidity and flow. Instead, the rigidity is set to constant, and the flow is computed using DiscreteFlow. Warning: This will severely impact performance!')
# GT data, if an evaluation is desired
parser.add_argument('--flow_fwd_gt', type=str, default='', help='GT flow in forward direction (used only for evaluation)')
parser.add_argument('--rigidity_gt', type=str, default='', help='GT rigidity (used only for evaluation)')
# Add algorithm parameters
parameters.add_parameters_argparse(parser)
# Parse all parameters
args = parser.parse_args()
if not args.rigidity:
vars(args)['rigidity_weight_cnn'] = 0.0
#
# Display all arguments
#
print('=== ARGUMENTS ===')
for k in sorted(vars(args).keys()):
v = vars(args)[k]
print('\t {} \t : \t {}'.format(k,v))
print('=================')
paths = {'path_image_prev': args.frame0,
'path_image_current': args.frame1,
'path_image_next': args.frame2,
'path_flow_fwd': args.flow_fwd,
'path_flow_bwd': args.flow_bwd,
'path_backflow_bwd': args.backflow_bwd,
'path_backflow_fwd': args.backflow_fwd,
'path_flow_fwd_gt': args.flow_fwd_gt,
'path_rigidity': args.rigidity,
'path_rigidity_gt': args.rigidity_gt}
# If tempdir does not exist yet, create it.
if not os.path.isdir(args.tempdir):
os.makedirs(args.tempdir)
# Call the computation with the provided paths
compute_mrflow(paths,args)
if __name__ == '__main__':
main()