forked from otoofim/ObjLocalisation
-
Notifications
You must be signed in to change notification settings - Fork 1
/
user_manual.html
612 lines (609 loc) · 59.4 KB
/
user_manual.html
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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8" />
<title>README</title>
<style>.markdown-preview:not([data-use-github-style]) { padding: 2em; font-size: 1.2em; color: rgb(171, 178, 191); background-color: rgb(40, 44, 52); overflow: auto; }
.markdown-preview:not([data-use-github-style]) > :first-child { margin-top: 0px; }
.markdown-preview:not([data-use-github-style]) h1, .markdown-preview:not([data-use-github-style]) h2, .markdown-preview:not([data-use-github-style]) h3, .markdown-preview:not([data-use-github-style]) h4, .markdown-preview:not([data-use-github-style]) h5, .markdown-preview:not([data-use-github-style]) h6 { line-height: 1.2; margin-top: 1.5em; margin-bottom: 0.5em; color: rgb(255, 255, 255); }
.markdown-preview:not([data-use-github-style]) h1 { font-size: 2.4em; font-weight: 300; }
.markdown-preview:not([data-use-github-style]) h2 { font-size: 1.8em; font-weight: 400; }
.markdown-preview:not([data-use-github-style]) h3 { font-size: 1.5em; font-weight: 500; }
.markdown-preview:not([data-use-github-style]) h4 { font-size: 1.2em; font-weight: 600; }
.markdown-preview:not([data-use-github-style]) h5 { font-size: 1.1em; font-weight: 600; }
.markdown-preview:not([data-use-github-style]) h6 { font-size: 1em; font-weight: 600; }
.markdown-preview:not([data-use-github-style]) strong { color: rgb(255, 255, 255); }
.markdown-preview:not([data-use-github-style]) del { color: rgb(124, 135, 156); }
.markdown-preview:not([data-use-github-style]) a, .markdown-preview:not([data-use-github-style]) a code { color: rgb(82, 139, 255); }
.markdown-preview:not([data-use-github-style]) img { max-width: 100%; }
.markdown-preview:not([data-use-github-style]) > p { margin-top: 0px; margin-bottom: 1.5em; }
.markdown-preview:not([data-use-github-style]) > ul, .markdown-preview:not([data-use-github-style]) > ol { margin-bottom: 1.5em; }
.markdown-preview:not([data-use-github-style]) blockquote { margin: 1.5em 0px; font-size: inherit; color: rgb(124, 135, 156); border-color: rgb(75, 83, 98); border-width: 4px; }
.markdown-preview:not([data-use-github-style]) hr { margin: 3em 0px; border-top: 2px dashed rgb(75, 83, 98); background: none; }
.markdown-preview:not([data-use-github-style]) table { margin: 1.5em 0px; }
.markdown-preview:not([data-use-github-style]) th { color: rgb(255, 255, 255); }
.markdown-preview:not([data-use-github-style]) th, .markdown-preview:not([data-use-github-style]) td { padding: 0.66em 1em; border: 1px solid rgb(75, 83, 98); }
.markdown-preview:not([data-use-github-style]) code { color: rgb(255, 255, 255); background-color: rgb(58, 63, 75); }
.markdown-preview:not([data-use-github-style]) pre.editor-colors { margin: 1.5em 0px; padding: 1em; font-size: 0.92em; border-radius: 3px; background-color: rgb(49, 54, 63); }
.markdown-preview:not([data-use-github-style]) kbd { color: rgb(255, 255, 255); border-width: 1px 1px 2px; border-style: solid; border-color: rgb(75, 83, 98) rgb(75, 83, 98) rgb(62, 68, 81); border-image: initial; background-color: rgb(58, 63, 75); }
.markdown-preview[data-use-github-style] { font-family: "Helvetica Neue", Helvetica, "Segoe UI", Arial, freesans, sans-serif; line-height: 1.6; word-wrap: break-word; padding: 30px; font-size: 16px; color: rgb(51, 51, 51); background-color: rgb(255, 255, 255); overflow: scroll; }
.markdown-preview[data-use-github-style] > :first-child { margin-top: 0px !important; }
.markdown-preview[data-use-github-style] > :last-child { margin-bottom: 0px !important; }
.markdown-preview[data-use-github-style] a:not([href]) { color: inherit; text-decoration: none; }
.markdown-preview[data-use-github-style] .absent { color: rgb(204, 0, 0); }
.markdown-preview[data-use-github-style] .anchor { position: absolute; top: 0px; left: 0px; display: block; padding-right: 6px; padding-left: 30px; margin-left: -30px; }
.markdown-preview[data-use-github-style] .anchor:focus { outline: none; }
.markdown-preview[data-use-github-style] h1, .markdown-preview[data-use-github-style] h2, .markdown-preview[data-use-github-style] h3, .markdown-preview[data-use-github-style] h4, .markdown-preview[data-use-github-style] h5, .markdown-preview[data-use-github-style] h6 { position: relative; margin-top: 1em; margin-bottom: 16px; font-weight: bold; line-height: 1.4; }
.markdown-preview[data-use-github-style] h1 .octicon-link, .markdown-preview[data-use-github-style] h2 .octicon-link, .markdown-preview[data-use-github-style] h3 .octicon-link, .markdown-preview[data-use-github-style] h4 .octicon-link, .markdown-preview[data-use-github-style] h5 .octicon-link, .markdown-preview[data-use-github-style] h6 .octicon-link { display: none; color: rgb(0, 0, 0); vertical-align: middle; }
.markdown-preview[data-use-github-style] h1:hover .anchor, .markdown-preview[data-use-github-style] h2:hover .anchor, .markdown-preview[data-use-github-style] h3:hover .anchor, .markdown-preview[data-use-github-style] h4:hover .anchor, .markdown-preview[data-use-github-style] h5:hover .anchor, .markdown-preview[data-use-github-style] h6:hover .anchor { padding-left: 8px; margin-left: -30px; text-decoration: none; }
.markdown-preview[data-use-github-style] h1:hover .anchor .octicon-link, .markdown-preview[data-use-github-style] h2:hover .anchor .octicon-link, .markdown-preview[data-use-github-style] h3:hover .anchor .octicon-link, .markdown-preview[data-use-github-style] h4:hover .anchor .octicon-link, .markdown-preview[data-use-github-style] h5:hover .anchor .octicon-link, .markdown-preview[data-use-github-style] h6:hover .anchor .octicon-link { display: inline-block; }
.markdown-preview[data-use-github-style] h1 tt, .markdown-preview[data-use-github-style] h2 tt, .markdown-preview[data-use-github-style] h3 tt, .markdown-preview[data-use-github-style] h4 tt, .markdown-preview[data-use-github-style] h5 tt, .markdown-preview[data-use-github-style] h6 tt, .markdown-preview[data-use-github-style] h1 code, .markdown-preview[data-use-github-style] h2 code, .markdown-preview[data-use-github-style] h3 code, .markdown-preview[data-use-github-style] h4 code, .markdown-preview[data-use-github-style] h5 code, .markdown-preview[data-use-github-style] h6 code { font-size: inherit; }
.markdown-preview[data-use-github-style] h1 { padding-bottom: 0.3em; font-size: 2.25em; line-height: 1.2; border-bottom: 1px solid rgb(238, 238, 238); }
.markdown-preview[data-use-github-style] h1 .anchor { line-height: 1; }
.markdown-preview[data-use-github-style] h2 { padding-bottom: 0.3em; font-size: 1.75em; line-height: 1.225; border-bottom: 1px solid rgb(238, 238, 238); }
.markdown-preview[data-use-github-style] h2 .anchor { line-height: 1; }
.markdown-preview[data-use-github-style] h3 { font-size: 1.5em; line-height: 1.43; }
.markdown-preview[data-use-github-style] h3 .anchor { line-height: 1.2; }
.markdown-preview[data-use-github-style] h4 { font-size: 1.25em; }
.markdown-preview[data-use-github-style] h4 .anchor { line-height: 1.2; }
.markdown-preview[data-use-github-style] h5 { font-size: 1em; }
.markdown-preview[data-use-github-style] h5 .anchor { line-height: 1.1; }
.markdown-preview[data-use-github-style] h6 { font-size: 1em; color: rgb(119, 119, 119); }
.markdown-preview[data-use-github-style] h6 .anchor { line-height: 1.1; }
.markdown-preview[data-use-github-style] p, .markdown-preview[data-use-github-style] blockquote, .markdown-preview[data-use-github-style] ul, .markdown-preview[data-use-github-style] ol, .markdown-preview[data-use-github-style] dl, .markdown-preview[data-use-github-style] table, .markdown-preview[data-use-github-style] pre { margin-top: 0px; margin-bottom: 16px; }
.markdown-preview[data-use-github-style] hr { height: 4px; padding: 0px; margin: 16px 0px; background-color: rgb(231, 231, 231); border: 0px none; }
.markdown-preview[data-use-github-style] ul, .markdown-preview[data-use-github-style] ol { padding-left: 2em; }
.markdown-preview[data-use-github-style] ul.no-list, .markdown-preview[data-use-github-style] ol.no-list { padding: 0px; list-style-type: none; }
.markdown-preview[data-use-github-style] ul ul, .markdown-preview[data-use-github-style] ul ol, .markdown-preview[data-use-github-style] ol ol, .markdown-preview[data-use-github-style] ol ul { margin-top: 0px; margin-bottom: 0px; }
.markdown-preview[data-use-github-style] li > p { margin-top: 16px; }
.markdown-preview[data-use-github-style] dl { padding: 0px; }
.markdown-preview[data-use-github-style] dl dt { padding: 0px; margin-top: 16px; font-size: 1em; font-style: italic; font-weight: bold; }
.markdown-preview[data-use-github-style] dl dd { padding: 0px 16px; margin-bottom: 16px; }
.markdown-preview[data-use-github-style] blockquote { padding: 0px 15px; color: rgb(119, 119, 119); border-left: 4px solid rgb(221, 221, 221); }
.markdown-preview[data-use-github-style] blockquote > :first-child { margin-top: 0px; }
.markdown-preview[data-use-github-style] blockquote > :last-child { margin-bottom: 0px; }
.markdown-preview[data-use-github-style] table { display: block; width: 100%; overflow: auto; word-break: keep-all; }
.markdown-preview[data-use-github-style] table th { font-weight: bold; }
.markdown-preview[data-use-github-style] table th, .markdown-preview[data-use-github-style] table td { padding: 6px 13px; border: 1px solid rgb(221, 221, 221); }
.markdown-preview[data-use-github-style] table tr { background-color: rgb(255, 255, 255); border-top: 1px solid rgb(204, 204, 204); }
.markdown-preview[data-use-github-style] table tr:nth-child(2n) { background-color: rgb(248, 248, 248); }
.markdown-preview[data-use-github-style] img { max-width: 100%; box-sizing: border-box; }
.markdown-preview[data-use-github-style] .emoji { max-width: none; }
.markdown-preview[data-use-github-style] span.frame { display: block; overflow: hidden; }
.markdown-preview[data-use-github-style] span.frame > span { display: block; float: left; width: auto; padding: 7px; margin: 13px 0px 0px; overflow: hidden; border: 1px solid rgb(221, 221, 221); }
.markdown-preview[data-use-github-style] span.frame span img { display: block; float: left; }
.markdown-preview[data-use-github-style] span.frame span span { display: block; padding: 5px 0px 0px; clear: both; color: rgb(51, 51, 51); }
.markdown-preview[data-use-github-style] span.align-center { display: block; overflow: hidden; clear: both; }
.markdown-preview[data-use-github-style] span.align-center > span { display: block; margin: 13px auto 0px; overflow: hidden; text-align: center; }
.markdown-preview[data-use-github-style] span.align-center span img { margin: 0px auto; text-align: center; }
.markdown-preview[data-use-github-style] span.align-right { display: block; overflow: hidden; clear: both; }
.markdown-preview[data-use-github-style] span.align-right > span { display: block; margin: 13px 0px 0px; overflow: hidden; text-align: right; }
.markdown-preview[data-use-github-style] span.align-right span img { margin: 0px; text-align: right; }
.markdown-preview[data-use-github-style] span.float-left { display: block; float: left; margin-right: 13px; overflow: hidden; }
.markdown-preview[data-use-github-style] span.float-left span { margin: 13px 0px 0px; }
.markdown-preview[data-use-github-style] span.float-right { display: block; float: right; margin-left: 13px; overflow: hidden; }
.markdown-preview[data-use-github-style] span.float-right > span { display: block; margin: 13px auto 0px; overflow: hidden; text-align: right; }
.markdown-preview[data-use-github-style] code, .markdown-preview[data-use-github-style] tt { padding: 0.2em 0px; margin: 0px; font-size: 85%; background-color: rgba(0, 0, 0, 0.04); border-radius: 3px; }
.markdown-preview[data-use-github-style] code::before, .markdown-preview[data-use-github-style] tt::before, .markdown-preview[data-use-github-style] code::after, .markdown-preview[data-use-github-style] tt::after { letter-spacing: -0.2em; content: " "; }
.markdown-preview[data-use-github-style] code br, .markdown-preview[data-use-github-style] tt br { display: none; }
.markdown-preview[data-use-github-style] del code { text-decoration: inherit; }
.markdown-preview[data-use-github-style] pre > code { padding: 0px; margin: 0px; font-size: 100%; word-break: normal; white-space: pre; background: transparent; border: 0px; }
.markdown-preview[data-use-github-style] .highlight { margin-bottom: 16px; }
.markdown-preview[data-use-github-style] .highlight pre, .markdown-preview[data-use-github-style] pre { padding: 16px; overflow: auto; font-size: 85%; line-height: 1.45; background-color: rgb(247, 247, 247); border-radius: 3px; }
.markdown-preview[data-use-github-style] .highlight pre { margin-bottom: 0px; word-break: normal; }
.markdown-preview[data-use-github-style] pre { word-wrap: normal; }
.markdown-preview[data-use-github-style] pre code, .markdown-preview[data-use-github-style] pre tt { display: inline; max-width: initial; padding: 0px; margin: 0px; overflow: initial; line-height: inherit; word-wrap: normal; background-color: transparent; border: 0px; }
.markdown-preview[data-use-github-style] pre code::before, .markdown-preview[data-use-github-style] pre tt::before, .markdown-preview[data-use-github-style] pre code::after, .markdown-preview[data-use-github-style] pre tt::after { content: normal; }
.markdown-preview[data-use-github-style] kbd { display: inline-block; padding: 3px 5px; font-size: 11px; line-height: 10px; color: rgb(85, 85, 85); vertical-align: middle; background-color: rgb(252, 252, 252); border-width: 1px; border-style: solid; border-color: rgb(204, 204, 204) rgb(204, 204, 204) rgb(187, 187, 187); border-image: initial; border-radius: 3px; box-shadow: rgb(187, 187, 187) 0px -1px 0px inset; }
.markdown-preview[data-use-github-style] a { color: rgb(51, 122, 183); }
.markdown-preview[data-use-github-style] code { color: inherit; }
.markdown-preview[data-use-github-style] pre.editor-colors { padding: 0.8em 1em; margin-bottom: 1em; font-size: 0.85em; border-radius: 4px; overflow: auto; }
.scrollbars-visible-always .markdown-preview pre.editor-colors .vertical-scrollbar, .scrollbars-visible-always .markdown-preview pre.editor-colors .horizontal-scrollbar { visibility: hidden; }
.scrollbars-visible-always .markdown-preview pre.editor-colors:hover .vertical-scrollbar, .scrollbars-visible-always .markdown-preview pre.editor-colors:hover .horizontal-scrollbar { visibility: visible; }
.markdown-preview .task-list-item-checkbox { position: absolute; margin: 0.25em 0px 0px -1.4em; }
.bracket-matcher .region {
border-bottom: 1px dotted lime;
position: absolute;
}
.line-number.bracket-matcher.bracket-matcher {
color: #abb2bf;
background-color: #3a3f4b;
}
.spell-check-misspelling .region {
border-bottom: 2px dotted rgba(255, 51, 51, 0.75);
}
.spell-check-corrections {
width: 25em !important;
}
pre.editor-colors {
background-color: #282c34;
color: #abb2bf;
}
pre.editor-colors .line.cursor-line {
background-color: rgba(153, 187, 255, 0.04);
}
pre.editor-colors .invisible {
color: #abb2bf;
}
pre.editor-colors .cursor {
border-left: 2px solid #528bff;
}
pre.editor-colors .selection .region {
background-color: #3e4451;
}
pre.editor-colors .bracket-matcher .region {
border-bottom: 1px solid #528bff;
box-sizing: border-box;
}
pre.editor-colors .invisible-character {
color: rgba(171, 178, 191, 0.15);
}
pre.editor-colors .indent-guide {
color: rgba(171, 178, 191, 0.15);
}
pre.editor-colors .wrap-guide {
background-color: rgba(171, 178, 191, 0.15);
}
pre.editor-colors .find-result .region.region.region,
pre.editor-colors .current-result .region.region.region {
border-radius: 2px;
background-color: rgba(82, 139, 255, 0.24);
transition: border-color 0.4s;
}
pre.editor-colors .find-result .region.region.region {
border: 2px solid transparent;
}
pre.editor-colors .current-result .region.region.region {
border: 2px solid #528bff;
transition-duration: .1s;
}
pre.editor-colors .gutter .line-number {
color: #636d83;
-webkit-font-smoothing: antialiased;
}
pre.editor-colors .gutter .line-number.cursor-line {
color: #abb2bf;
background-color: #3a3f4b;
}
pre.editor-colors .gutter .line-number.cursor-line-no-selection {
background-color: transparent;
}
pre.editor-colors .gutter .line-number .icon-right {
color: #abb2bf;
}
pre.editor-colors .gutter:not(.git-diff-icon) .line-number.git-line-removed.git-line-removed::before {
bottom: -3px;
}
pre.editor-colors .gutter:not(.git-diff-icon) .line-number.git-line-removed::after {
content: "";
position: absolute;
left: 0px;
bottom: 0px;
width: 25px;
border-bottom: 1px dotted rgba(224, 82, 82, 0.5);
pointer-events: none;
}
pre.editor-colors .gutter .line-number.folded,
pre.editor-colors .gutter .line-number:after,
pre.editor-colors .fold-marker:after {
color: #abb2bf;
}
.syntax--comment {
color: #5c6370;
font-style: italic;
}
.syntax--comment .syntax--markup.syntax--link {
color: #5c6370;
}
.syntax--entity.syntax--name.syntax--type {
color: #e5c07b;
}
.syntax--entity.syntax--other.syntax--inherited-class {
color: #98c379;
}
.syntax--keyword {
color: #c678dd;
}
.syntax--keyword.syntax--control {
color: #c678dd;
}
.syntax--keyword.syntax--operator {
color: #abb2bf;
}
.syntax--keyword.syntax--other.syntax--special-method {
color: #61afef;
}
.syntax--keyword.syntax--other.syntax--unit {
color: #d19a66;
}
.syntax--storage {
color: #c678dd;
}
.syntax--storage.syntax--type.syntax--annotation,
.syntax--storage.syntax--type.syntax--primitive {
color: #c678dd;
}
.syntax--storage.syntax--modifier.syntax--package,
.syntax--storage.syntax--modifier.syntax--import {
color: #abb2bf;
}
.syntax--constant {
color: #d19a66;
}
.syntax--constant.syntax--variable {
color: #d19a66;
}
.syntax--constant.syntax--character.syntax--escape {
color: #56b6c2;
}
.syntax--constant.syntax--numeric {
color: #d19a66;
}
.syntax--constant.syntax--other.syntax--color {
color: #56b6c2;
}
.syntax--constant.syntax--other.syntax--symbol {
color: #56b6c2;
}
.syntax--variable {
color: #e06c75;
}
.syntax--variable.syntax--interpolation {
color: #be5046;
}
.syntax--variable.syntax--parameter {
color: #abb2bf;
}
.syntax--string {
color: #98c379;
}
.syntax--string > .syntax--source,
.syntax--string .syntax--embedded {
color: #abb2bf;
}
.syntax--string.syntax--regexp {
color: #56b6c2;
}
.syntax--string.syntax--regexp .syntax--source.syntax--ruby.syntax--embedded {
color: #e5c07b;
}
.syntax--string.syntax--other.syntax--link {
color: #e06c75;
}
.syntax--punctuation.syntax--definition.syntax--comment {
color: #5c6370;
}
.syntax--punctuation.syntax--definition.syntax--method-parameters,
.syntax--punctuation.syntax--definition.syntax--function-parameters,
.syntax--punctuation.syntax--definition.syntax--parameters,
.syntax--punctuation.syntax--definition.syntax--separator,
.syntax--punctuation.syntax--definition.syntax--seperator,
.syntax--punctuation.syntax--definition.syntax--array {
color: #abb2bf;
}
.syntax--punctuation.syntax--definition.syntax--heading,
.syntax--punctuation.syntax--definition.syntax--identity {
color: #61afef;
}
.syntax--punctuation.syntax--definition.syntax--bold {
color: #e5c07b;
font-weight: bold;
}
.syntax--punctuation.syntax--definition.syntax--italic {
color: #c678dd;
font-style: italic;
}
.syntax--punctuation.syntax--section.syntax--embedded {
color: #be5046;
}
.syntax--punctuation.syntax--section.syntax--method,
.syntax--punctuation.syntax--section.syntax--class,
.syntax--punctuation.syntax--section.syntax--inner-class {
color: #abb2bf;
}
.syntax--support.syntax--class {
color: #e5c07b;
}
.syntax--support.syntax--type {
color: #56b6c2;
}
.syntax--support.syntax--function {
color: #56b6c2;
}
.syntax--support.syntax--function.syntax--any-method {
color: #61afef;
}
.syntax--entity.syntax--name.syntax--function {
color: #61afef;
}
.syntax--entity.syntax--name.syntax--class,
.syntax--entity.syntax--name.syntax--type.syntax--class {
color: #e5c07b;
}
.syntax--entity.syntax--name.syntax--section {
color: #61afef;
}
.syntax--entity.syntax--name.syntax--tag {
color: #e06c75;
}
.syntax--entity.syntax--other.syntax--attribute-name {
color: #d19a66;
}
.syntax--entity.syntax--other.syntax--attribute-name.syntax--id {
color: #61afef;
}
.syntax--meta.syntax--class {
color: #e5c07b;
}
.syntax--meta.syntax--class.syntax--body {
color: #abb2bf;
}
.syntax--meta.syntax--method-call,
.syntax--meta.syntax--method {
color: #abb2bf;
}
.syntax--meta.syntax--definition.syntax--variable {
color: #e06c75;
}
.syntax--meta.syntax--link {
color: #d19a66;
}
.syntax--meta.syntax--require {
color: #61afef;
}
.syntax--meta.syntax--selector {
color: #c678dd;
}
.syntax--meta.syntax--separator {
color: #abb2bf;
}
.syntax--meta.syntax--tag {
color: #abb2bf;
}
.syntax--underline {
text-decoration: underline;
}
.syntax--none {
color: #abb2bf;
}
.syntax--invalid.syntax--deprecated {
color: #523d14 !important;
background-color: #e0c285 !important;
}
.syntax--invalid.syntax--illegal {
color: white !important;
background-color: #e05252 !important;
}
.syntax--markup.syntax--bold {
color: #d19a66;
font-weight: bold;
}
.syntax--markup.syntax--changed {
color: #c678dd;
}
.syntax--markup.syntax--deleted {
color: #e06c75;
}
.syntax--markup.syntax--italic {
color: #c678dd;
font-style: italic;
}
.syntax--markup.syntax--heading {
color: #e06c75;
}
.syntax--markup.syntax--heading .syntax--punctuation.syntax--definition.syntax--heading {
color: #61afef;
}
.syntax--markup.syntax--link {
color: #56b6c2;
}
.syntax--markup.syntax--inserted {
color: #98c379;
}
.syntax--markup.syntax--quote {
color: #d19a66;
}
.syntax--markup.syntax--raw {
color: #98c379;
}
.syntax--source.syntax--c .syntax--keyword.syntax--operator {
color: #c678dd;
}
.syntax--source.syntax--cpp .syntax--keyword.syntax--operator {
color: #c678dd;
}
.syntax--source.syntax--cs .syntax--keyword.syntax--operator {
color: #c678dd;
}
.syntax--source.syntax--css .syntax--property-name,
.syntax--source.syntax--css .syntax--property-value {
color: #828997;
}
.syntax--source.syntax--css .syntax--property-name.syntax--support,
.syntax--source.syntax--css .syntax--property-value.syntax--support {
color: #abb2bf;
}
.syntax--source.syntax--elixir .syntax--source.syntax--embedded.syntax--source {
color: #abb2bf;
}
.syntax--source.syntax--elixir .syntax--constant.syntax--language,
.syntax--source.syntax--elixir .syntax--constant.syntax--numeric,
.syntax--source.syntax--elixir .syntax--constant.syntax--definition {
color: #61afef;
}
.syntax--source.syntax--elixir .syntax--variable.syntax--definition,
.syntax--source.syntax--elixir .syntax--variable.syntax--anonymous {
color: #c678dd;
}
.syntax--source.syntax--elixir .syntax--parameter.syntax--variable.syntax--function {
color: #d19a66;
font-style: italic;
}
.syntax--source.syntax--elixir .syntax--quoted {
color: #98c379;
}
.syntax--source.syntax--elixir .syntax--keyword.syntax--special-method,
.syntax--source.syntax--elixir .syntax--embedded.syntax--section,
.syntax--source.syntax--elixir .syntax--embedded.syntax--source.syntax--empty {
color: #e06c75;
}
.syntax--source.syntax--elixir .syntax--readwrite.syntax--module .syntax--punctuation {
color: #e06c75;
}
.syntax--source.syntax--elixir .syntax--regexp.syntax--section,
.syntax--source.syntax--elixir .syntax--regexp.syntax--string {
color: #be5046;
}
.syntax--source.syntax--elixir .syntax--separator,
.syntax--source.syntax--elixir .syntax--keyword.syntax--operator {
color: #d19a66;
}
.syntax--source.syntax--elixir .syntax--variable.syntax--constant {
color: #e5c07b;
}
.syntax--source.syntax--elixir .syntax--array,
.syntax--source.syntax--elixir .syntax--scope,
.syntax--source.syntax--elixir .syntax--section {
color: #828997;
}
.syntax--source.syntax--gfm .syntax--markup {
-webkit-font-smoothing: auto;
}
.syntax--source.syntax--gfm .syntax--link .syntax--entity {
color: #61afef;
}
.syntax--source.syntax--go .syntax--storage.syntax--type.syntax--string {
color: #c678dd;
}
.syntax--source.syntax--ini .syntax--keyword.syntax--other.syntax--definition.syntax--ini {
color: #e06c75;
}
.syntax--source.syntax--java .syntax--storage.syntax--modifier.syntax--import {
color: #e5c07b;
}
.syntax--source.syntax--java .syntax--storage.syntax--type {
color: #e5c07b;
}
.syntax--source.syntax--java .syntax--keyword.syntax--operator.syntax--instanceof {
color: #c678dd;
}
.syntax--source.syntax--java-properties .syntax--meta.syntax--key-pair {
color: #e06c75;
}
.syntax--source.syntax--java-properties .syntax--meta.syntax--key-pair > .syntax--punctuation {
color: #abb2bf;
}
.syntax--source.syntax--js .syntax--keyword.syntax--operator {
color: #56b6c2;
}
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--delete,
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--in,
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--of,
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--instanceof,
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--new,
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--typeof,
.syntax--source.syntax--js .syntax--keyword.syntax--operator.syntax--void {
color: #c678dd;
}
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--dictionary.syntax--json > .syntax--string.syntax--quoted.syntax--json {
color: #e06c75;
}
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--dictionary.syntax--json > .syntax--string.syntax--quoted.syntax--json > .syntax--punctuation.syntax--string {
color: #e06c75;
}
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--dictionary.syntax--json > .syntax--value.syntax--json > .syntax--string.syntax--quoted.syntax--json,
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--array.syntax--json > .syntax--value.syntax--json > .syntax--string.syntax--quoted.syntax--json,
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--dictionary.syntax--json > .syntax--value.syntax--json > .syntax--string.syntax--quoted.syntax--json > .syntax--punctuation,
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--array.syntax--json > .syntax--value.syntax--json > .syntax--string.syntax--quoted.syntax--json > .syntax--punctuation {
color: #98c379;
}
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--dictionary.syntax--json > .syntax--constant.syntax--language.syntax--json,
.syntax--source.syntax--json .syntax--meta.syntax--structure.syntax--array.syntax--json > .syntax--constant.syntax--language.syntax--json {
color: #56b6c2;
}
.syntax--ng.syntax--interpolation {
color: #e06c75;
}
.syntax--ng.syntax--interpolation.syntax--begin,
.syntax--ng.syntax--interpolation.syntax--end {
color: #61afef;
}
.syntax--ng.syntax--interpolation .syntax--function {
color: #e06c75;
}
.syntax--ng.syntax--interpolation .syntax--function.syntax--begin,
.syntax--ng.syntax--interpolation .syntax--function.syntax--end {
color: #61afef;
}
.syntax--ng.syntax--interpolation .syntax--bool {
color: #d19a66;
}
.syntax--ng.syntax--interpolation .syntax--bracket {
color: #abb2bf;
}
.syntax--ng.syntax--pipe,
.syntax--ng.syntax--operator {
color: #abb2bf;
}
.syntax--ng.syntax--tag {
color: #56b6c2;
}
.syntax--ng.syntax--attribute-with-value .syntax--attribute-name {
color: #e5c07b;
}
.syntax--ng.syntax--attribute-with-value .syntax--string {
color: #c678dd;
}
.syntax--ng.syntax--attribute-with-value .syntax--string.syntax--begin,
.syntax--ng.syntax--attribute-with-value .syntax--string.syntax--end {
color: #abb2bf;
}
.syntax--source.syntax--ruby .syntax--constant.syntax--other.syntax--symbol > .syntax--punctuation {
color: inherit;
}
.syntax--source.syntax--php .syntax--class.syntax--bracket {
color: #abb2bf;
}
.syntax--source.syntax--python .syntax--keyword.syntax--operator.syntax--logical.syntax--python {
color: #c678dd;
}
.syntax--source.syntax--python .syntax--variable.syntax--parameter {
color: #d19a66;
}
</style>
</head>
<body class='markdown-preview' data-use-github-style><h1 id="objlocalisation">ObjLocalisation</h1>
<p>The aim of this project is to learn a policy to localize objects in images by turning visual attention to the salient parts of images. In order to achieve this goal, the popular RL algorithm, Q-learning, is adopted by incorporating the approximation method, CNNs. <a href="https://www.nature.com/articles/nature14236">DQL</a> is the method resulting from cooperating Q-learning and CNNs. While using this method for object localization is not new and was tried before in <a href="https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Caicedo_Active_Object_Localization_ICCV_2015_paper.pdf">Active Object Localization with Deep Reinforcement Learning</a>, in this project despite implementing the algorithm with the novel deep learning framework, Tensorflow, a new set of experiments were conducted by using a new neural network architecture to show that representation learning can happen by Q-learning. More specifically, the original paper uses a pre-trained CNN as a feature extractor. However, in this project, the model was trained without using a pre-trained network for feature extraction. This MSc project was conducted in <a href="https://www.gla.ac.uk/schools/computing/research/researchoverview/computervisionandautonomoussystems/">Computer Vision & Autonomous Systems Group</a> at the university of Glasgow under supervision of <a href="http://www.dcs.gla.ac.uk/~psiebert/">Dr Jan Paul Siebert</a>. Below is the examples of a trained model on VOC 2012 dataset. The following sections describe user manual for researhers who intend to use this implementation. In addition, in order to make modifying the implementation modest all the files are commented.</p>
<div align="center" class="row">
<div align="center" class="column">
<img width="900px" src="https://drive.google.com/uc?export=view&id=1QsOi-zVPicMfMej0OfFBV52cDQKGJbEh">
</div>
</div>
<h1 id="getting-started">Getting started</h1>
<p>You can clone this project using this <a href="https://github.com/otoofim/ObjLocalisation.git">link</a> and install requierments by <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">pip install -r requirements.txt</code>. The <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">requirements.txt</code> would install everything you need. However, before using the <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">requirements.txt</code>, it is suggested to create a virtual environment with python 2.7.12. This code was developed and tested on Ubuntu 16.04 using Python 2.7.12 and Tensorflow 1.8. The code works fine for Tensorflow 1.8 however, in order to run the code on the Glasgow university cluster it requires some changes. In the file <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">DNN.py</code> the parts of the code that need to be modified in order to run on GPU cluster is marked as "Old API". In this way the code can run with older Tensorflow APIs. Otherwise, it is recommended to follow <a href="https://www.tensorflow.org/install/install_linux">this tutorial</a> to create a virtual environment and then install Tensorflow and all requirements. </p>
<h2 id="inputs">Inputs</h2>
<p>In this project <a href="http://host.robots.ox.ac.uk/pascal/VOC/voc2012/">Pascal VOC 2012</a> dataset was used to train the model. It is organized to download and prepare the dataset for training in the first run. However, if you need to use another dataset then the input pipeline needs to be modified. To change the default dataset you need to make some changes in the files <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">VOC2012DataProvider.py</code>, <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">readingFileEfficiently.py</code>, and <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">VOC2012_npz_files_writter.py</code>. In Pacal VOC dataset the gorund truth is provided seperately in <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">xml</code> files. For this resean, it is needed to write images and their corrensponding ground truth to a single file, .npz, in order to create image batches for efficient learning. That is done by <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">VOC2012_npz_files_writter.py</code>. Later .npz files are used by <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">DQL.py</code> for training. Since Pascal VOC 2012 consists of 19386 images loading all images into memory makes trouble. For this, <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">readingFileEfficiently.py</code> loads input images into memory in an efficient way. Further, <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">VOC2012DataProvider.py</code> reads .npz files and provides datapoints to <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">readingFileEfficiently.py</code>.</p>
<h2 id="command-line-options-and-configuration">Command Line Options and Configuration</h2>
<p>Having set up the environment, training can begin using <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">run_training.py</code>. Its command line options is as follow:</p>
<pre class="editor-colors lang-"><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> usage: run_training.py [-h] [-n NUM_EPISODES] [-rms REPLAY_MEMORY_SIZE]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-rmis REPLAY_MEMORY_INIT_SIZE]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-u UPDATE_TARGET_ESTIMATOR_EVERY] [-d DISCOUNT_FACTOR]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-es EPSILON_START] [-ee EPSILON_END]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-ed EPSILON_DECAY_STEPS] [-c CATEGORY [CATEGORY ...]]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-m MODEL_NAME]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Train an object localizer</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> optional arguments:</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -h, --help show this help message and exit</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -n NUM_EPISODES, --num_episodes NUM_EPISODES</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Number of episodes that the agect can interact with an</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> image. Default: 5</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -rms REPLAY_MEMORY_SIZE, --replay_memory_size REPLAY_MEMORY_SIZE</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Number of the most recent experiences that would be</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> stored. Default: 500000</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -rmis REPLAY_MEMORY_INIT_SIZE, --replay_memory_init_size REPLAY_MEMORY_INIT_SIZE</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Number of experiences to initialize replay memory.</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Default: 500</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -u UPDATE_TARGET_ESTIMATOR_EVERY, --update_target_estimator_every UPDATE_TARGET_ESTIMATOR_EVERY</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Number of steps after which estimator parameters are</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> copied to target network. Default: 10000</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -d DISCOUNT_FACTOR, --discount_factor DISCOUNT_FACTOR</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Discount factor. Default: 0.99</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -es EPSILON_START, --epsilon_start EPSILON_START</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Epsilon decay schedule start point. Default: 1.0</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -ee EPSILON_END, --epsilon_end EPSILON_END</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Epsilon decay schedule end point. Default: 0.2</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -ed EPSILON_DECAY_STEPS, --epsilon_decay_steps EPSILON_DECAY_STEPS</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Epsilon decay step rate. This number indicates epsilon</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> would be decearsed from start to end point after how</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> many steps. Default: 500</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -c CATEGORY [CATEGORY ...], --category CATEGORY [CATEGORY ...]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Indicating the categories are going to be used for</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> training. You can list name of the classes you want to</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> use in training. If you wish to use all classes then</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> you can use *. For instnce <-c cat dog>. Default: cat</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -m MODEL_NAME, --model_name MODEL_NAME</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> The trained model would be saved with this name under</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> the path ../experiments/model_name. Default:</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> default_model</span></span></div></pre><p>Note: If you need to train a model on multiple categories the command would be <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">python run_training.py -c cat dog</code>. In addition, if you want to trian a new mdoel on top of a previously trained model then you need to copy the content of the bestModel folder of the previously trained model to its checkpoints folder. In this way, the best model will be loaded for training. </p>
<p>To evaluate a trained model on the test set <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">run_testing.py</code> is used. Testing conditions can be set as below:</p>
<pre class="editor-colors lang-"><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>usage: run_testing.py [-h] [-n NUM_EPISODES] [-c CATEGORY [CATEGORY ...]]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-m MODEL_NAME]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>Evaluate a model on test set</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>optional arguments:</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -h, --help show this help message and exit</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -n NUM_EPISODES, --num_episodes NUM_EPISODES</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Number of episodes that the agent can interact with an</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> image. Default: 15</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -c CATEGORY [CATEGORY ...], --category CATEGORY [CATEGORY ...]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Indicating the categories are going to be used for</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> testing. You can list name of the classes you want to</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> use in testing, for instnce <-c cat dog>. If you wish</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> to use all classes then you can use *. Default: cat</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -m MODEL_NAME, --model_name MODEL_NAME</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> The model name that will be loaded for evaluation. Do</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> not forget to put the model under the path</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> ../experiments/model_name. Default: default_model</span></span></div></pre><p>There are two other python files that are useful for visualization purposes. <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">run_visulazing_actions.py</code> can be used to visualize a sequence of actions:</p>
<pre class="editor-colors lang-"><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>usage: run_visulazing_actions.py [-h] [-m MODEL_NAME] [-i IMAGE_PATH]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-g GROUND_TRUTH [GROUND_TRUTH ...]]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-n NAME]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>Visualizing sequence of actions</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>optional arguments:</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -h, --help show this help message and exit</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -m MODEL_NAME, --model_name MODEL_NAME</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> The model parameters that will be loaded for testing.</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Do not forget to put the model under the path</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> ../experiments/model_name. Default: default_model</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -i IMAGE_PATH, --image_path IMAGE_PATH</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Path to an image.</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -g GROUND_TRUTH [GROUND_TRUTH ...], --ground_truth GROUND_TRUTH [GROUND_TRUTH ...]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Target coordinates. The order of coordinates should be</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> like: xmin ymin xmax ymax. Default: 0 0 1 1</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -n NAME, --name NAME Name of the output file. It will be stored in</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> ../experiments/model_name/anim/</span></span></div></pre><p>In addition, the neural network layers can be visualized using <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">run_visulazing_layers.py</code>:</p>
<pre class="editor-colors lang-"><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>usage: run_visulazing_layers.py [-h] [-m MODEL_NAME] [-i IMAGE_PATH]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> [-ln LAYER_NUM]</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>Visualizing CNN layers</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> </span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span>optional arguments:</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -h, --help show this help message and exit</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -m MODEL_NAME, --model_name MODEL_NAME</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> The model parameters that will be loaded for testing.</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Do not forget to put the model under the path</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> ../experiments/model_name. Default: default_model</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -i IMAGE_PATH, --image_path IMAGE_PATH</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Path to an image.</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> -ln LAYER_NUM, --layer_num LAYER_NUM</span></span></div><div class="line"><span class="syntax--text syntax--plain syntax--null-grammar"><span> Layer number you wish to visualize.</span></span></div></pre><p><strong>Note:</strong> In all visualization and evaluation files the best model saved in the directory of the given model is used. </p>
<h2 id="outputs">Outputs</h2>
<p><strong>run_training.py:</strong></p>
<p> The output of trainng process is stored in <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">../experiments/ModelName</code>. The result will be saved in four folders. The first one is <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">summaries_q_estimator</code> consists of an TF event record. Using tensorboard, graphs related to the training can be visualised. To run tensorboard, it is needed to call tensorboard in this way <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">tensorboard --logdir=../experiments/Modelname/summaries_q_estimator</code>. The second folder is report. This folder includes <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">log.txt</code> file which is the log showed in terminal during training. The third one is 'checkpoints' folder contains three files, which corresponds to the final model saved at the end of training process. The final folder is 'bestModel' that includes the best model based on validation accuracy.</p>
<p><strong>run_testing.py:</strong></p>
<p>The output of evaluation process is stored in <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">../experiments/ModelName/report/evaluate_[categories].txt</code>. That file consists of the results evaluated separately on each category and mean average precision (MAP) over all categories.</p>
<p><strong>run_visulazing_actions.py:</strong></p>
<p>The output of this file is a short video shows the agent interactions with the given image. The result is stored in <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">../experiments/ModelName/anim</code>. </p>
<p><strong>run_visulazing_layers.py:</strong></p>
<p>The output of this script is a set of images each of which corresponds to a filter in a given layer. The result is stored in <code style="font-family: Menlo, Consolas, "DejaVu Sans Mono", monospace;">../experiments/ModelName/visu</code>. </p>
<h1 id="acknowledgments">Acknowledgments</h1>
<p>This code is implemented by getting help from the following sources:</p>
<ul>
<li><a href="https://github.com/jccaicedo/localization-agent">Original implementation of active object localization algorithm</a></li>
<li><a href="https://github.com/dennybritz/reinforcement-learning">Tutorial for deep reinforcement learning</a></li>
<li><a href="https://github.com/otoofim/mlpractical">Tutorial for deep learning from the university of Edinburgh</a></li>
</ul></body>
</html>