-
Notifications
You must be signed in to change notification settings - Fork 22
/
util.py
executable file
·436 lines (376 loc) · 18.6 KB
/
util.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
# ---------------------------------------------------------------------
# Copyright (c) 2018 TU Berlin, Communication Systems Group
# Written by Tobias Senst <[email protected]>
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
# ---------------------------------------------------------------------
import numpy as np
import cv2
import copy
import pickle
import math
def parameter_to_string(parameter_dict):
out_str = str()
for key in sorted(parameter_dict.keys()):
try:
str_value = str(parameter_dict[key])
except TypeError:
str_value = parameter_dict[key]
out_str = out_str + str_value + "_"
return out_str
def create_config(parameter, filelist):
config_list = list()
for id, f in enumerate(filelist):
file_list = copy.deepcopy(f)
config_list.append({ "files" : file_list, "parameter" : copy.deepcopy(parameter), "file_index": id})
return config_list
def computeEE(src0, src1):
diff_flow = src0 - src1
res = (diff_flow[:, :, 0] * diff_flow[:, :, 0]) + (diff_flow[:, :, 1] * diff_flow[:, :, 1])
return cv2.sqrt(res)
def computer_errors(ee_base, mask):
ee_base = ee_base * mask
ret, R1 = cv2.threshold(src=ee_base, thresh=1, maxval=1, type=cv2.THRESH_BINARY)
ret, R2 = cv2.threshold(src=ee_base, thresh=2, maxval=1, type=cv2.THRESH_BINARY)
ret, R3 = cv2.threshold(src=ee_base, thresh=3, maxval=1, type=cv2.THRESH_BINARY)
R3 = R3 * mask
R2 = R2 * mask
R1 = R1 * mask
r3_sum = cv2.sumElems(R3)[0]
r2_sum = cv2.sumElems(R2)[0]
r1_sum = cv2.sumElems(R1)[0]
ee = cv2.sumElems(ee_base)[0]
no_p = cv2.sumElems(mask)[0]
result = {"ee" : ee, "R1": r1_sum, "R2": r2_sum, "R3" : r3_sum, "noPoints" : no_p}
return result
def compute_error(flow, gt_flow, invalid_mask):
mag_flow = cv2.sqrt(gt_flow[:, :, 0] * gt_flow[:, :, 0] + gt_flow[:, :, 1] * gt_flow[:, :, 1])
ret, mask_to_large = cv2.threshold(src=mag_flow, thresh=900, maxval=1, type=cv2.THRESH_BINARY_INV)
total_inp_mask = invalid_mask[:, :, 0] + invalid_mask[:, :, 1] + invalid_mask[:, :, 2]
ret, fg_mask = cv2.threshold(src=invalid_mask[:, :, 1], thresh=0.5, maxval=1,
type=cv2.THRESH_BINARY)
ret, total_mask = cv2.threshold(src=total_inp_mask, thresh=0.5, maxval=1,
type=cv2.THRESH_BINARY)
#mask_to_large = np.ones(fg_mask.shape)
bg_mask = total_mask - fg_mask
ee_base = computeEE(flow, gt_flow)
result = dict()
result["FG"] = computer_errors(ee_base, fg_mask * mask_to_large)
result["BG"] = computer_errors(ee_base, bg_mask * mask_to_large)
result["Total"] = computer_errors(ee_base, total_mask * mask_to_large)
return result
def draw_trajectories(gt_trajectory, estimate_trajectory, img):
h = img.shape[0]
w = img.shape[1]
for n in range(len(gt_trajectory)):
pos = (int(gt_trajectory[n][0]),int(gt_trajectory[n][1]))
if pos[0] >= 0 and pos[0] < h and pos[1] >= 0 and pos[1] < w:
img[pos[0], pos[1],:] = [0,255,0]
#cv2.circle(img,center=(int(gt_trajectory[n][0]),int(gt_trajectory[n][1])),
# radius=1, color=(0,255,0), thickness=-1)
for n in range(0,len(estimate_trajectory),2):
pos = (int(estimate_trajectory[n]), int(estimate_trajectory[n+1]))
if pos[0] >= 0 and pos[0] < h and pos[1] >= 0 and pos[1] < w:
img[pos[0], pos[1], :] = [255, 0, 0]
#cv2.circle(img, center=(int(estimate_trajectory[n]),int(estimate_trajectory[n+1])),
# radius=2, color=(255,0,0), thickness=-1)
return img
def differenz_trajectory_list(gt_trajectories, estimate_trajectories):
"""
.@brief gt_trajectories and estimate trajectories have to be aligned
"""
differenz_trajectory_list = list()
assert len(gt_trajectories) == len(estimate_trajectories)
for n in range(len(gt_trajectories)):
if len(gt_trajectories[n]) != (len(estimate_trajectories[n])) / 2:
print( "ID", n, len(gt_trajectories[n]), (len(estimate_trajectories[n])) / 2)
for i in range(len(gt_trajectories[n])):
diff_x = gt_trajectories[n][i][0] - estimate_trajectories[n][2*i]
diff_y = gt_trajectories[n][i][1] - estimate_trajectories[n][2*i+1]
differenz_trajectory_list.append(math.sqrt( diff_x * diff_x + diff_y * diff_y))
return np.array(differenz_trajectory_list)
def compute_tracking_error(differenz_trajectory_list, thresholds_list):
result = list()
for thresholds in thresholds_list:
ret, mask = cv2.threshold(src=differenz_trajectory_list, thresh=thresholds, maxval=1, type=cv2.THRESH_BINARY_INV)
result.append(cv2.sumElems(mask)[0] / differenz_trajectory_list.shape[0])
return result
def get_trajectory_lengths(groundtruth_trajectory_list):
result = list()
for item in groundtruth_trajectory_list:
result.append(len(item))
return result
def flow2RGB(flow, max_flow_mag = 5):
""" Color-coded visualization of optical flow fields
# Arguments
flow: array of shape [:,:,2] containing optical flow
max_flow_mag: maximal expected flow magnitude used to normalize. If max_flow_mag < 0 the maximal
magnitude of the optical flow field will be used
"""
hsv_mat = np.ones(shape=(flow.shape[0], flow.shape[1], 3), dtype=np.float32) * 255
ee = cv2.sqrt(flow[:, :, 0] * flow[:, :, 0] + flow[:, :, 1] * flow[:, :, 1])
angle = np.arccos(flow[:, :, 0]/ ee)
angle[flow[:, :, 0] == 0] = 0
angle[flow[:, :, 1] == 0] = 6.2831853 - angle[flow[:, :, 1] == 0]
angle = angle * 180 / 3.141
hsv_mat[:,:,0] = angle
if max_flow_mag < 0:
max_flow_mag = ee.max()
hsv_mat[:,:,1] = ee * 255.0 / max_flow_mag
ret, hsv_mat[:,:,1] = cv2.threshold(src=hsv_mat[:,:,1], maxval=255, thresh=255, type=cv2.THRESH_TRUNC )
rgb_mat = cv2.cvtColor(hsv_mat.astype(np.uint8), cv2.COLOR_HSV2BGR)
return rgb_mat
def readFlowFiles(filename):
with open(filename, 'rb') as f:
magic = np.fromfile(f, np.float32, count=1)
if 202021.25 != magic:
print( "Magic number incorrect. Invalid .flo file")
else:
w = np.fromfile(f, np.int32, count=1)[0]
h = np.fromfile(f, np.int32, count=1)[0]
data = np.fromfile(f, np.float32, count=2 * w * h)
# Reshape data into 3D array (columns, rows, bands)
data2D = np.resize(data, (h, w, 2))
return data2D
def writeFlowFile(filename,flow):
TAG_STRING = np.array(202021.25, dtype=np.float32)
assert flow.shape[2] == 2
h = np.array(flow.shape[0], dtype=np.int32)
w = np.array(flow.shape[1], dtype=np.int32)
with open(filename, 'wb') as f:
f.write(TAG_STRING.tobytes())
f.write(w.tobytes())
f.write(h.tobytes())
f.write(flow.tobytes())
def flow2RGB(flow, max_flow_mag = 5):
""" Color-coded visualization of optical flow fields
# Arguments
flow: array of shape [:,:,2] containing optical flow
max_flow_mag: maximal expected flow magnitude used to normalize. If max_flow_mag < 0 the maximal
magnitude of the optical flow field will be used
"""
hsv_mat = np.ones(shape=(flow.shape[0], flow.shape[1], 3), dtype=np.float32) * 255
ee = cv2.sqrt(flow[:, :, 0] * flow[:, :, 0] + flow[:, :, 1] * flow[:, :, 1])
angle = np.arccos(flow[:, :, 0]/ ee)
angle[flow[:, :, 0] == 0] = 0
angle[flow[:, :, 1] == 0] = 6.2831853 - angle[flow[:, :, 1] == 0]
angle = angle * 180 / 3.141
hsv_mat[:,:,0] = angle
if max_flow_mag < 0:
max_flow_mag = ee.max()
hsv_mat[:,:,1] = ee * 220.0 / max_flow_mag
ret, hsv_mat[:,:,1] = cv2.threshold(src=hsv_mat[:,:,1], maxval=255, thresh=255, type=cv2.THRESH_TRUNC )
rgb_mat = cv2.cvtColor(hsv_mat.astype(np.uint8), cv2.COLOR_HSV2BGR)
return rgb_mat
def drawFlowField(filename, flow):
cv2.imwrite(filename=filename, img = flow2RGB(flow))
def avg_sequence(src):
sequence_result = dict()
for seq_keys in src.keys():
result = dict()
for item in src[seq_keys]:
for key in item.keys():
if key != "FG" and key != "BG" and key != "Total" :
continue
if key not in result:
result[key] = item[key]
else:
for key1 in item[key].keys():
result[key][key1] += item[key][key1]
for key in result.keys():
result[key]["ee"] = result[key]["ee"] / result[key]["noPoints"]
result[key]["R1"] = result[key]["R1"] / result[key]["noPoints"]
result[key]["R2"] = result[key]["R2"] / result[key]["noPoints"]
result[key]["R3"] = result[key]["R3"] / result[key]["noPoints"]
sequence_result[seq_keys] = copy.deepcopy(result)
return sequence_result
def get_sequence_measures(result_list):
sequence_list = dict()
for item in result_list:
parameter_str = parameter_to_string(item[0]["parameter"])
if "dir" in item[0]["files"]:
seq_name = item[0]["files"]["dir"]
else:
seq_name = "None"
if seq_name.find("_dyn") >= 0 :
continue
if parameter_str not in sequence_list:
sequence_list[parameter_str] = dict()
if seq_name not in sequence_list[parameter_str]:
sequence_list[parameter_str][seq_name] = list()
sequence_list[parameter_str][seq_name].append(item[1])
return sequence_list
def avg_measures(src):
total_result = dict()
for seq_keys in src.keys():
result = dict()
for item in src[seq_keys]:
for key in item.keys():
if key == "time":
continue
if key not in result:
result[key] = item[key]
else:
for key1 in item[key].keys():
result[key][key1] += int(item[key][key1])
for key in result.keys():
result[key]["ee"] = result[key]["ee"] / result[key]["noPoints"]
result[key]["R1"] = result[key]["R1"] / result[key]["noPoints"]
result[key]["R2"] = result[key]["R2"] / result[key]["noPoints"]
result[key]["R3"] = result[key]["R3"] / result[key]["noPoints"]
if len(total_result) == 0:
for key in result.keys():
total_result[key] = dict()
total_result[key]["ee"] = result[key]["ee"] / len(src)
total_result[key]["R1"] = result[key]["R1"] / len(src)
total_result[key]["R2"] = result[key]["R2"] / len(src)
total_result[key]["R3"] = result[key]["R3"] / len(src)
else:
for key in result.keys():
total_result[key]["ee"] += result[key]["ee"] / len(src)
total_result[key]["R1"] += result[key]["R1"] / len(src)
total_result[key]["R2"] += result[key]["R2"] / len(src)
total_result[key]["R3"] += result[key]["R3"] / len(src)
return total_result
def avg_measures_no_dict(src):
total_result = dict()
for seq_keys in src.keys():
result = dict()
for item in src[seq_keys]:
for key in item.keys():
if key not in result:
result[key] = item[key]
else:
result[key] += item[key]
result["ee"] = result["ee"] / result["no_points"]
result["R1"] = result["R1"] / result["no_points"]
result["R2"] = result["R2"] / result["no_points"]
result["R3"] = result["R3"] / result["no_points"]
if len(total_result) == 0:
for key in result.keys():
total_result = dict()
total_result["ee"] = result["ee"] / len(src)
total_result["R1"] = result["R1"] / len(src)
total_result["R2"] = result["R2"] / len(src)
total_result["R3"] = result["R3"] / len(src)
else:
for key in result.keys():
total_result["ee"] += result["ee"] / len(src)
total_result["R1"] += result["R1"] / len(src)
total_result["R2"] += result["R2"] / len(src)
total_result["R3"] += result["R3"] / len(src)
return total_result
def avg_sequences(sequence_list, use_type):
res_FG_ee = []
res_FG_R2 = []
res_BG_ee = []
res_BG_R2 = []
res_total_ee = []
res_total_R2 = []
for seq_name in sequence_list.keys():
if use_type == 1 and seq_name.find("_hDyn") == -1:
continue
if use_type == 0 and seq_name.find("_hDyn") >= 0 :
continue
res_FG_ee.append(sequence_list[seq_name]["FG"]["ee"])
res_FG_R2.append(sequence_list[seq_name]["FG"]["R2"])
res_BG_ee.append(sequence_list[seq_name]["BG"]["ee"])
res_BG_R2.append(sequence_list[seq_name]["BG"]["R2"])
res_total_ee.append(sequence_list[seq_name]["Total"]["ee"])
res_total_R2.append(sequence_list[seq_name]["Total"]["R2"])
return np.mean(res_FG_ee), 100 * np.mean(res_FG_R2), \
np.mean(res_BG_ee), 100 * np.mean(res_BG_R2), \
np.mean(res_total_ee), 100 * np.mean(res_total_R2)
def getLatexTable(filename):
data = pickle.load(open(filename, "rb"))
str_result = "\\begin{table} \n \\centering " \
"\\begin{tabular}{l|crcr|crcr|crcrcr|r} \n" \
"\\hline \n " \
"\\multicolumn{1}{c|}{} & \\multicolumn{2}{|c}{FG (Static) } & \\multicolumn{2}{c|} { BG (Static)} & " \
"\\multicolumn{2}{|c}{FG (Dynamic)} & \\multicolumn{2}{c|}{ BG (Dynamic)} & " \
"\\multicolumn{2}{c}{FG($\\varnothing$)}&\multicolumn{2}{c}{BG ($\\varnothing$)} & " \
"\\multicolumn{2}{c|}{$\\varnothing$} \\\\ \n " \
"\\multicolumn{1}{c|}{}& EPE & R2[\\%] & EPE & R2[\\%]& EPE & R2[\\%]& EPE & R2[\\%]" \
"& EPE & R2[\\%]& EPE & R2[\\%]& EPE & R2[\\%] \\\\ \n "
result_list = data["result"]
method_result_list = get_sequence_measures(result_list)
for method_key in method_result_list.keys():
sequence_result = avg_sequence(method_result_list[method_key])
# print(sequence_result)
ret_static = avg_sequences(sequence_result, 0)
ret_dynamic = avg_sequences(sequence_result, 1)
ret_total = avg_sequences(sequence_result, 2)
name = method_key.replace("/","")
name = name.replace("_", "")
str_out = method_key \
+ " & {:.3f}".format(ret_static[0]) \
+ " & {:.2f}".format(ret_static[1]) \
+ " & {:.3f}".format(ret_static[2]) \
+ " & {:.2f}".format(ret_static[3]) \
+ " & {:.3f}".format(ret_dynamic[0]) \
+ " & {:.2f}".format(ret_dynamic[1]) \
+ " & {:.3f}".format(ret_dynamic[2]) \
+ " & {:.2f}".format(ret_dynamic[3]) \
+ " & {:.3f}".format(ret_total[0]) \
+ " & {:.3f}".format(ret_total[1]) \
+ " & {:.3f}".format(ret_total[2]) \
+ " & {:.2f}".format(ret_total[3]) \
+ " & {:.3f}".format(ret_total[4]) \
+ " & {:.2f}".format(ret_total[5]) \
+ " \\\ "
str_result = str_result + str_out + "\n"
str_result = str_result + "\end{tabular} \n " \
"\\vspace{0.1cm} \n" \
"\\caption{Evaluation results common optical flow metrics. " \
"Dynamic comprised sequences with and static without camera motion, " \
"BG - background motion vectors and FG - motion vectors located at persons of the crowd.} \n" \
"\\end{table}"
return str_result
def genTrajectoryLatexTable(filename, item_key = "dense_person"):
str_result = "\n \\begin{table} \n " \
"\\scriptsize \n " \
"\\setlength{\\tabcolsep}{2.4pt} \n " \
"\\centering \n " \
"\\begin{tabular}{l|cc|cc|cc|cc|cc|c} \n " \
" & \\multicolumn{2}{c|}{IM01 (Dyn)} & \\multicolumn{2}{c|}{IM02 (Dyn)} " \
" & \\multicolumn{2}{c|}{IM03 (Dyn)} & \\multicolumn{2}{c|}{IM04 (Dyn)} " \
" & \\multicolumn{2}{c|}{IM05 (Dyn)} & $\\varnothing$ \\\\ \n" \
"\\hline \n"
methods = pickle.load(open(filename, "rb"))
for ret in methods:
result = ret["name"] \
+ " & " + "{:2.2f}".format(100 * ret["IM01"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM01_hDyn"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM02"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM02_hDyn"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM03"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM03_hDyn"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM04"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM04_hDyn"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM05"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["IM05_hDyn"][item_key][14]) \
+ " & " + "{:2.2f}".format(100 * ret["all"][item_key][14]) + "\\\ "
str_result = str_result + result + "\n"
if item_key == "person":
name_str = "person trajectories"
else:
name_str = "dense person trajectories"
str_result = str_result + "\\hline \n " \
"\\end{tabular} \n " \
"\\vspace{0.01cm} \n " \
"\\caption{Evaluation results with long-term motion metric (" + name_str + ")" \
"The \\textbf{tracking accuracy} in percentage for the threshold set to 15 pixels. " \
"Higher values denote more accurate results.} \n " \
"\\end{table} \n"
return str_result