forked from opencv/opencv
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathhaartraining.htm
676 lines (518 loc) · 33 KB
/
haartraining.htm
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
<html>
<head>
<meta http-equiv=Content-Type content="text/html; charset=windows-1251">
<meta name=Generator content="Microsoft Word 11 (filtered)">
<title>Object Detection Using Haar-like Features with Cascade of Boosted
Classifiers</title>
<style>
<!--
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0in;
margin-bottom:.0001pt;
text-align:justify;
font-size:12.0pt;
font-family:"Times New Roman";}
h1
{margin-top:12.0pt;
margin-right:0in;
margin-bottom:3.0pt;
margin-left:0in;
text-align:justify;
page-break-after:avoid;
font-size:16.0pt;
font-family:Arial;}
h2
{margin-top:12.0pt;
margin-right:0in;
margin-bottom:3.0pt;
margin-left:0in;
text-align:justify;
page-break-after:avoid;
font-size:14.0pt;
font-family:Arial;
font-style:italic;}
h3
{margin-top:12.0pt;
margin-right:0in;
margin-bottom:3.0pt;
margin-left:0in;
text-align:justify;
page-break-after:avoid;
font-size:13.0pt;
font-family:Arial;}
span.Typewch
{font-family:"Courier New";
font-weight:bold;}
@page Section1
{size:595.3pt 841.9pt;
margin:56.7pt 88.0pt 63.2pt 85.05pt;}
div.Section1
{page:Section1;}
/* List Definitions */
ol
{margin-bottom:0in;}
ul
{margin-bottom:0in;}
-->
</style>
</head>
<body lang=RU>
<div class=Section1>
<h1><span lang=EN-US>Rapid Object Detection With A Cascade of Boosted
Classifiers Based on Haar-like Features</span></h1>
<h2><span lang=EN-US>Introduction</span></h2>
<p class=MsoNormal><span lang=EN-US>This document describes how to train and
use a cascade of boosted classifiers for rapid object detection. A large set of
over-complete haar-like features provide the basis for the simple individual
classifiers. Examples of object detection tasks are face, eye and nose
detection, as well as logo detection. </span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>The sample detection task in this document
is logo detection, since logo detection does not require the collection of
large set of registered and carefully marked object samples. Instead we assume
that from one prototype image, a very large set of derived object examples can
be derived (</span><span class=Typewch><span lang=EN-US>createsamples</span></span><span
lang=EN-US> utility, see below).</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>A detailed description of the training/evaluation
algorithm can be found in [1] and [2].</span></p>
<h2><span lang=EN-US>Samples Creation</span></h2>
<p class=MsoNormal><span lang=EN-US>For training a training samples must be
collected. There are two sample types: negative samples and positive samples.
Negative samples correspond to non-object images. Positive samples correspond
to object images.</span></p>
<h3><span lang=EN-US>Negative Samples</span></h3>
<p class=MsoNormal><span lang=EN-US>Negative samples are taken from arbitrary
images. These images must not contain object representations. Negative samples
are passed through background description file. It is a text file in which each
text line contains the filename (relative to the directory of the description
file) of negative sample image. This file must be created manually. Note that
the negative samples and sample images are also called background samples or
background samples images, and are used interchangeably in this document</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Example of negative description file:</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Directory structure:</span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>/img</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US> img1.jpg</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US> img2.jpg</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>bg.txt</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US> </span></span></p>
<p class=MsoNormal><span class=Typewch><span style='font-family:"Times New Roman";
font-weight:normal'>File </span></span><span class=Typewch><span lang=EN-US>bg.txt:</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>img/img1.jpg</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>img/img2.jpg</span></span></p>
<h3><span lang=EN-US>Positive Samples</span></h3>
<p class=MsoNormal><span lang=EN-US>Positive samples are created by </span><span
class=Typewch><span lang=EN-US>createsamples</span></span><span lang=EN-US>
utility. They may be created from single object image or from collection of
previously marked up images.<br>
<br>
</span></p>
<p class=MsoNormal><span lang=EN-US>The single object image may for instance
contain a company logo. Then are large set of positive samples are created from
the given object image by randomly rotating, changing the logo color as well as
placing the logo on arbitrary background.</span></p>
<p class=MsoNormal><span lang=EN-US>The amount and range of randomness can be
controlled by command line arguments. </span></p>
<p class=MsoNormal><span lang=EN-US>Command line arguments:</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- vec <vec_file_name></span></span><span
lang=EN-US> </span></p>
<p class=MsoNormal style='margin-left:17.1pt'><span lang=EN-US>name of the
output file containing the positive samples for training</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- img <image_file_name></span></span><span
lang=EN-US> </span></p>
<p class=MsoNormal style='margin-left:17.1pt'><span lang=EN-US>source object
image (e.g., a company logo)</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- bg <background_file_name></span></span><span
lang=EN-US> </span></p>
<p class=MsoNormal style='margin-left:17.1pt'><span lang=EN-US>background
description file; contains a list of images into which randomly distorted
versions of the object are pasted for positive sample generation</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- num <number_of_samples></span></span><span
lang=EN-US> </span></p>
<p class=MsoNormal style='margin-left:17.1pt'><span lang=EN-US>number of
positive samples to generate </span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- bgcolor <background_color></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
lang=EN-US> background color (currently grayscale images are assumed); the
background color denotes the transparent color. Since there might be
compression artifacts, the amount of color tolerance can be specified by </span><span
class=Typewch><span lang=EN-US>–bgthresh</span></span><span class=Typewch><span
lang=EN-US style='font-family:Arial;font-weight:normal'>. </span></span><span
lang=EN-US>All pixels between </span><span class=Typewch><span lang=EN-US>bgcolor-bgthresh</span></span><span
lang=EN-US> and </span><span class=Typewch><span lang=EN-US>bgcolor+bgthresh</span></span><span
lang=EN-US> are regarded as transparent.</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- bgthresh <background_color_threshold></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- inv</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
lang=EN-US> if specified, the colors will be inverted</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- randinv</span></span><span lang=EN-US> </span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
lang=EN-US> if specified, the colors will be inverted randomly</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxidev <max_intensity_deviation></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US> </span></span><span lang=EN-US>maximal
intensity deviation of foreground samples pixels</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxxangle <max_x_rotation_angle>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxyangle <max_y_rotation_angle>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxzangle <max_z_rotation_angle></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
lang=EN-US> maximum rotation angles in radians</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>-show</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
lang=EN-US> if specified, each sample will be shown. Pressing ‘Esc’ will
continue creation process without samples showing. Useful debugging option.</span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- w <sample_width></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US> </span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'>width (in
pixels) of the output samples</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- h <sample_height></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US> </span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'>height (in
pixels) of the output samples</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US> </span></span></p>
<p class=MsoNormal><span lang=EN-US>For following procedure is used to create a
sample object instance:</span></p>
<p class=MsoNormal><span lang=EN-US>The source image is rotated random around
all three axes. The chosen angle is limited my</span><span class=Typewch><span
lang=EN-US> -max?angle</span></span><span lang=EN-US>. Next pixels of
intensities in the range of </span><span class=Typewch><span lang=EN-US>[bg_color-bg_color_threshold;
bg_color+bg_color_threshold]</span></span><span lang=EN-US> are regarded as
transparent. White noise is added to the intensities of the foreground. If </span><span
class=Typewch><span lang=EN-US>–inv</span></span><span lang=EN-US> key is
specified then foreground pixel intensities are inverted. If </span><span
class=Typewch><span lang=EN-US>–randinv</span></span><span lang=EN-US> key is
specified then it is randomly selected whether for this sample inversion will
be applied. Finally, the obtained image is placed onto arbitrary background
from the background description file, resized to the pixel size specified by </span><span
class=Typewch><span lang=EN-US>–w</span></span><span lang=EN-US> and </span><span
class=Typewch><span lang=EN-US>–h</span></span><span lang=EN-US> and stored
into the file specified by the </span><span class=Typewch><span lang=EN-US>–vec</span></span><span
lang=EN-US> command line parameter.</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Positive samples also may be obtained from
a collection of previously marked up images. This collection is described by
text file similar to background description file. Each line of this file
corresponds to collection image. The first element of the line is image file
name. It is followed by number of object instances. The following numbers are
the coordinates of bounding rectangles (x, y, width, height).</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Example of description file:</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Directory structure:</span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>/img</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US> img1.jpg</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US> img2.jpg</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>info.dat</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US> </span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US style='font-family:
"Times New Roman";font-weight:normal'>File </span></span><span class=Typewch><span
lang=EN-US>info.dat:</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>img/img1.jpg 1 140
100 45 45</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>img/img2.jpg 2 100
200 50 50 50 30 25 25</span></span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Image </span><span class=Typewch><span
lang=EN-US>img1.jpg</span></span><span lang=EN-US> contains single object
instance with bounding rectangle (140, 100, 45, 45). Image </span><span
class=Typewch><span lang=EN-US>img2.jpg</span></span><span lang=EN-US> contains
two object instances.</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>In order to create positive samples from
such collection </span><span class=Typewch><span lang=EN-US>–info</span></span><span
lang=EN-US> argument should be specified instead of </span><span class=Typewch><span
lang=EN-US>–img</span></span><span class=Typewch><span style='font-family:"Times New Roman";
font-weight:normal'>:</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- info <collection_file_name></span></span><span
lang=EN-US> </span></p>
<p class=MsoNormal style='margin-left:17.1pt'><span lang=EN-US>description file
of marked up images collection</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>The scheme of sample creation in this case
is as follows. The object instances are taken from images. Then they are
resized to samples size and stored in output file. No distortion is applied, so
the only affecting arguments are </span><span class=Typewch><span lang=EN-US>–w</span></span><span
lang=EN-US>, </span><span class=Typewch><span lang=EN-US>-h</span></span><span
lang=EN-US>, </span><span class=Typewch><span lang=EN-US>-show</span></span><span
lang=EN-US> and </span><span class=Typewch><span lang=EN-US>–num</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>.</span></span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>createsamples</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> utility may be used for examining samples stored in positive samples
file. In order to do this only </span></span><span class=Typewch><span
lang=EN-US>–vec</span></span><span class=Typewch><span lang=EN-US
style='font-family:"Times New Roman";font-weight:normal'>, </span></span><span
class=Typewch><span lang=EN-US>–w</span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'> and </span></span><span
class=Typewch><span lang=EN-US>–h</span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'> parameters
should be specified.</span></span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Note that for training, it does not matter
how positive samples files are generated. So the </span><span class=Typewch><span
lang=EN-US>createsamples</span></span><span lang=EN-US> utility is only one way
to collect/create a vector file of positive samples.</span></p>
<h2><span lang=EN-US>Training</span></h2>
<p class=MsoNormal><span lang=EN-US>The next step after samples creation is
training of classifier. It is performed by the </span><span class=Typewch><span
lang=EN-US>haartraining</span></span><span lang=EN-US> utility.</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Command line arguments:</span><span
class=Typewch><span lang=EN-US> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- data <dir_name></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> directory name in which the trained classifier is stored</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- vec <vec_file_name></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> file name of positive sample file (created by </span></span><span
class=Typewch><span lang=EN-US>trainingsamples</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> utility or by any other means)</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- bg <background_file_name></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> background description file</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- npos <number_of_positive_samples>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- nneg <number_of_negative_samples></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> number of positive/negative samples used in training of each
classifier stage. Reasonable values are npos = 7000 and nneg = 3000.</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- nstages <number_of_stages></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US> </span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'>number of
stages to be trained</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- nsplits <number_of_splits></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> determines the weak classifier used in stage classifiers. If </span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman"'>1</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>, then a simple stump classifier is used, if </span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman"'>2</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> and more, then CART classifier with </span></span><span class=Typewch><span
lang=EN-US>number_of_splits</span></span><span class=Typewch><span lang=EN-US
style='font-family:"Times New Roman";font-weight:normal'> internal (split)
nodes is used</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- mem <memory_in_MB></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> Available memory in MB for precalculation. The more memory you
have the faster the training process</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- sym (default),</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- nonsym</span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> specifies whether the object class under training has vertical
symmetry or not. Vertical symmetry speeds up training process. For instance,
frontal faces show off vertical symmetry</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- minhitrate <min_hit_rate></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> minimal desired hit rate for each stage classifier. Overall hit
rate may be estimated as </span></span><span class=Typewch><span lang=EN-US>(min_hit_rate^number_of_stages)</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxfalsealarm <max_false_alarm_rate></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> maximal desired false alarm rate for each stage classifier. </span></span><span
class=Typewch><span style='font-family:"Times New Roman";font-weight:normal'>Overall
false alarm rate may be estimated as</span></span><span class=Typewch><span
lang=EN-US> (max_false_alarm_rate^number_of_stages)</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- weighttrimming <weight_trimming></span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US> </span></span><span class=Typewch><span
lang=EN-US style='font-family:"Times New Roman";font-weight:normal'>Specifies
whether and how much weight trimming should be used. A decent choice is 0.90.</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- eqw</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- mode <BASIC (default) | CORE | ALL></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> selects the type of haar features set used in training. BASIC use
only upright features, while ALL uses the full set of upright and 45 degree
rotated feature set. See [1] for more details.</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- w <sample_width>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- h <sample_height></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> Size of training samples (in pixels). Must have exactly the same
values as used during training samples creation (utility </span></span><span
class=Typewch><span lang=EN-US>trainingsamples</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>)</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US style='font-family:
"Times New Roman";font-weight:normal'> </span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US style='font-family:
"Times New Roman";font-weight:normal'>Note: in order to use multiprocessor
advantage a compiler that supports OpenMP 1.0 standard should be used.</span></span></p>
<h2><span lang=EN-US>Application</span></h2>
<p class=MsoNormal><span lang=EN-US>OpenCV cvHaarDetectObjects() function (in
particular haarFaceDetect demo) is used for detection.</span></p>
<h3><span lang=EN-US>Test Samples</span></h3>
<p class=MsoNormal><span lang=EN-US>In order to evaluate the performance of
trained classifier a collection of marked up images is needed. When such
collection is not available test samples may be created from single object
image by </span><span class=Typewch><span lang=EN-US>createsamples</span></span><span
lang=EN-US> utility. The scheme of test samples creation in this case is
similar to training samples creation since each test sample is a background
image into which a randomly distorted and randomly scaled instance of the
object picture is pasted at a random position. </span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>If both </span><span class=Typewch><span
lang=EN-US>–img</span></span><span lang=EN-US> and </span><span class=Typewch><span
lang=EN-US>–info</span></span><span lang=EN-US> arguments are specified then
test samples will be created by </span><span class=Typewch><span lang=EN-US>createsamples</span></span><span
lang=EN-US> utility. The sample image is arbitrary distorted as it was
described below, then it is placed at random location to background image and
stored. The corresponding description line is added to the file specified by </span><span
class=Typewch><span lang=EN-US>–info</span></span><span lang=EN-US> argument.</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>The </span><span class=Typewch><span
lang=EN-US>–w</span></span><span lang=EN-US> and </span><span class=Typewch><span
lang=EN-US>–h</span></span><span lang=EN-US> keys determine the minimal size of
placed object picture.</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>The test image file name format is as
follows:</span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US>imageOrderNumber_x_y_width_height.jpg</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>, where </span></span><span class=Typewch><span lang=EN-US>x</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>, </span></span><span class=Typewch><span lang=EN-US>y</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>, </span></span><span class=Typewch><span lang=EN-US>width</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> and </span></span><span class=Typewch><span lang=EN-US>height</span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> are the coordinates of placed object bounding rectangle.</span></span></p>
<p class=MsoNormal><span class=Typewch><span lang=EN-US style='font-family:
"Times New Roman";font-weight:normal'>Note that you should use a background
images set different from the background image set used during training.</span></span></p>
<h3><span class=Typewch><span lang=EN-US style='font-family:"Times New Roman"'>Performance
Evaluation</span></span></h3>
<p class=MsoNormal><span lang=EN-US>In order to evaluate the performance of the
classifier </span><span class=Typewch><span lang=EN-US>performance</span></span><span
lang=EN-US> utility may be used. It takes a collection of marked up images,
applies the classifier and outputs the performance, i.e. number of found
objects, number of missed objects, number of false alarms and other
information.</span></p>
<p class=MsoNormal><span lang=EN-US> </span></p>
<p class=MsoNormal><span lang=EN-US>Command line arguments:</span><span
class=Typewch><span lang=EN-US> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- data <dir_name></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> directory name in which the trained classifier is stored</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- info <collection_file_name></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> file with test samples description</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxSizeDiff <max_size_difference></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- maxPosDiff <max_position_difference></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> determine the criterion of reference and detected rectangles
coincidence. Default values are 1.5 and 0.3 respectively.</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- sf <scale_factor></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> detection parameter. Default value is 1.2.</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- w <sample_width>,</span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US>- h <sample_height></span></span><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> </span></span></p>
<p class=MsoNormal style='margin-left:17.1pt;text-indent:-17.1pt'><span
class=Typewch><span lang=EN-US style='font-family:"Times New Roman";font-weight:
normal'> Size of training samples (in pixels). Must have exactly the same
values as used during training (utility </span></span><span class=Typewch><span
lang=EN-US>haartraining</span></span><span class=Typewch><span lang=EN-US
style='font-family:"Times New Roman";font-weight:normal'>)</span></span></p>
<h2><span lang=EN-US>References</span></h2>
<p class=MsoNormal><span lang=EN-US>[1] Rainer Lienhart and Jochen Maydt. An
Extended Set of Haar-like Features for Rapid Object Detection. Submitted to
ICIP2002.</span></p>
<p class=MsoNormal><span lang=EN-US>[2] Alexander Kuranov, Rainer Lienhart, and
Vadim Pisarevsky. An Empirical Analysis of Boosting Algorithms for Rapid
Objects With an Extended Set of Haar-like Features. Intel Technical Report
MRL-TR-July02-01, 2002.</span></p>
</div>
</body>
</html>