CVPR CVPR

#***** #*****
CVPR 2023 Submission #*****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE.

000 054
001 055
002 056
003 LATEX Author Guidelines for CVPR Proceedings 057
004 058
005 059
006 060
Anonymous CVPR submission
007 061
008 062
009 Paper ID ***** 063
010 064
011 065
012 066
Abstract such papers will not be reviewed is that there is no provi-
013 067
sion for supervised revisions of manuscripts. The review-
014 068
The ABSTRACT is to be in fully justified italicized text, ing process cannot determine the suitability of the paper for
015 069
at the top of the left-hand column, below the author and presentation in eight pages if it is reviewed in eleven.
016 070
affiliation information. Use the word “Abstract” as the title,
017 1.4. The ruler 071
in 12-point Times, boldface type, centered relative to the
018 072
column, initially capitalized. The abstract is to be in 10- The LATEX style defines a printed ruler which should be
019 073
point, single-spaced type. Leave two blank lines after the present in the version submitted for review. The ruler is
020 074
Abstract, then begin the main text. Look at previous CVPR provided in order that reviewers may comment on partic-
021 075
abstracts to get a feel for style and length. ular lines in the paper without circumlocution. If you are
022 076
preparing a document using a non-LATEX document prepara-
023 077
tion system, please arrange for an equivalent ruler to appear
024 078
1. Introduction on the final output pages. The presence or absence of the
025 079
ruler should not change the appearance of any other content
026 Please follow the steps outlined below when submitting on the page. The camera-ready copy should not contain a 080
027 your manuscript to the IEEE Computer Society Press. This ruler. (LATEX users may use options of cvpr.sty to switch 081
028 style guide now has several important modifications (for ex- between different versions.) 082
029 ample, you are no longer warned against the use of sticky Reviewers: note that the ruler measurements do not align 083
030 tape to attach your artwork to the paper), so all authors well with lines in the paper — this turns out to be very dif- 084
031 should read this new version. ficult to do well when the paper contains many figures and 085
032 086
equations, and, when done, looks ugly. Just use fractional
033 1.1. Language 087
references (e.g., this line is 087.5), although in most cases
034 088
All manuscripts must be in English. one would expect that the approximate location will be ad-
035 089
equate.
036 1.2. Dual submission 090
037 1.5. Paper ID 091
038 Please refer to the author guidelines on the CVPR 2023 092
web page for a discussion of the policy on dual submissions. Make sure that the Paper ID from the submission sys-
039 093
tem is visible in the version submitted for review (replacing
040 094
1.3. Paper length the “*****” you see in this document). If you are using
041 095
the LATEX template, make sure to update paper ID in the
042 Papers, excluding the references section, must be no 096
appropriate place in the tex file.
043 longer than eight pages in length. The references section 097
044 will not be included in the page count, and there is no limit 1.6. Mathematics 098
045 on the length of the references section. For example, a pa- 099
Please number all of your sections and displayed equa-
046 per of eight pages with two pages of references would have 100
tions as in these examples:
047 a total length of 10 pages. There will be no extra page 101
048 charges for CVPR 2023. E = m · c2 (1) 102
049 Overlength papers will simply not be reviewed. This in- 103
050 cludes papers where the margins and formatting are deemed and 104
051 to have been significantly altered from those laid down by v = a · t. (2) 105
052 this style guide. Note that this LATEX guide already sets fig- It is important for readers to be able to refer to any particu- 106
053 ure captions and references in a smaller font. The reason lar equation. Just because you did not refer to it in the text 107

1
CVPR CVPR
#***** #*****
CVPR 2023 Submission #*****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE.

108 162
does not mean some future reader might not need to refer to [1] Authors. “The frobnicatable foo filter”, F&G
109 163
it. It is cumbersome to have to use circumlocutions like “the 2014 Submission ID 324, Supplied as supplemen-
110 164
equation second from the top of page 3 column 1”. (Note tal material fg324.pdf.
111 165
that the ruler will not be present in the final copy, so is not
112
an alternative to equation numbers). All authors will benefit Finally, you may feel you need to tell the reader that 166
113
from reading Mermin’s description of how to write math- more details can be found elsewhere, and refer them to 167
114
ematics: http://www.pamitc.org/documents/ a technical report. For conference submissions, the pa- 168
115
mermin.pdf. per must stand on its own, and not require the reviewer 169
116 to go to a tech report for further details. Thus, you may 170
117 1.7. Blind review say in the body of the paper “further details may be found 171
118 in [?]”. Then submit the tech report as supplemental ma- 172
Many authors misunderstand the concept of anonymiz-
119 terial. Again, you may not assume the reviewers will read 173
ing for blind review. Blind review does not mean that one
120 this material. 174
must remove citations to one’s own work—in fact it is often
121 Sometimes your paper is about a problem which you 175
impossible to review a paper unless the previous citations
122 tested using a tool that is widely known to be restricted to a 176
are known and available.
123 single institution. For example, let’s say it’s 1969, you have 177
Blind review means that you do not use the words “my”
124 solved a key problem on the Apollo lander, and you believe 178
or “our” when citing previous work. That is all. (But see
125 that the CVPR70 audience would like to hear about your 179
below for tech reports.)
126 solution. The work is a development of your celebrated 180
Saying “this builds on the work of Lucy Smith [1]” does
127 1968 paper entitled “Zero-g frobnication: How being the 181
not say that you are Lucy Smith; it says that you are building
128 only people in the world with access to the Apollo lander 182
on her work. If you are Smith and Jones, do not say “as we
129 source code makes us a wow at parties”, by Zeus et al. 183
130
show in [7]”, say “as Smith and Jones show in [7]” and at 184
You can handle this paper like any other. Do not write
131
the end of the paper, include reference 7 as you would any 185
“We show how to improve our previous work [Anonymous,
132
other cited work. 186
1968]. This time we tested the algorithm on a lunar lander
133 An example of a bad paper just asking to be rejected: [name of lander removed for blind review]”. That would be 187
134 silly, and would immediately identify the authors. Instead 188
An analysis of the frobnicatable foo filter.
135 write the following: 189
136 In this paper we present a performance analysis 190
137 of our previous paper [1], and show it to be in- We describe a system for zero-g frobnication. 191
138 ferior to all previously known methods. Why the This system is new because it handles the fol- 192
139 previous paper was accepted without this analysis lowing cases: A, B. Previous systems [Zeus et al. 193
140 is beyond me. 1968] did not handle case B properly. Ours han- 194
141 [1] Removed for blind review dles it by including a foo term in the bar integral. 195
142 ... 196
143 An example of an acceptable paper: The proposed system was integrated with the 197
144 Apollo lunar lander, and went all the way to the 198
An analysis of the frobnicatable foo filter. moon, don’t you know. It displayed the follow-
145 199
146 In this paper we present a performance analysis of ing behaviours, which show how well we solved 200
147 the paper of Smith et al. [1], and show it to be in- cases A and B: ... 201
148 ferior to all previously known methods. Why the 202
As you can see, the above text follows standard scientific
149 previous paper was accepted without this analysis 203
convention, reads better than the first version, and does not
150 is beyond me. 204
explicitly name you as the authors. A reviewer might think
151 [1] Smith, L and Jones, C. “The frobnicatable 205
it likely that the new paper was written by Zeus et al., but
152 foo filter, a fundamental contribution to human 206
cannot make any decision based on that guess. He or she
153 knowledge”. Nature 381(12), 1-213. 207
would have to be sure that no other authors could have been
154 208
contracted to solve problem B.
155 If you are making a submission to another conference at 209
156 the same time, which covers similar or overlapping mate- FAQ 210
157 rial, you may need to refer to that submission in order to 211
Q: Are acknowledgements OK?
158 explain the differences, just as you would if you had pre- 212
A: No. Leave them for the final copy.
159 viously published related work. In such cases, include the 213
160 anonymized parallel submission [?] as supplemental mate- Q: How do I cite my results reported in open challenges? 214
161 rial and cite it as A: To conform with the double-blind review policy, you 215

2
CVPR CVPR
#***** #*****
CVPR 2023 Submission #*****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE.

216 270
main title (on the first page) should begin 1 inch (2.54 cm)
217 271
from the top edge of the page. The second and following
218 272
pages should begin 1 inch (2.54 cm) from the top edge. On
219 273
all pages, the bottom margin should be 1 18 inches (2.86 cm)
220 274
from the bottom edge of the page for 8.5 × 11-inch paper;
221 275
for A4 paper, approximately 1 58 inches (4.13 cm) from the
222 276
bottom edge of the page.
223 277
224 2.1. Margins and page numbering 278
225 279
226
All printed material, including text, illustrations, and 280
227
charts, must be kept within a print area 6 78 inches (17.46 281
228
cm) wide by 8 87 inches (22.54 cm) high. Page numbers 282
229
should be in the footer, centered and 34 inches from the bot- 283
230
Figure 1. Example of caption. It is set in Roman so that mathemat- tom of the page. The review version should have page num- 284
ics (always set in Roman: B sin A = A sin B) may be included bers, yet the final version submitted as camera ready should
231 285
without an ugly clash. not show any page numbers. The LATEX template takes care
232 286
233
of this when used properly. 287
234 288
can report results of other challenge participants together 2.2. Type style and fonts
235 with your results in your paper. For your results, however, 289
236 Wherever Times is specified, Times Roman may also be 290
you should not identify yourself and should not mention
237 used. If neither is available on your word processor, please 291
your participation in the challenge. Instead present your
238 use the font closest in appearance to Times to which you 292
results referring to the method proposed in your paper and
239 have access. 293
draw conclusions based on the experimental comparison to
240 MAIN TITLE. Center the title 1 83 inches (3.49 cm) from 294
other results.
241 the top edge of the first page. The title should be in Times 295
242 14-point, boldface type. Capitalize the first letter of nouns, 296
243 pronouns, verbs, adjectives, and adverbs; do not capitalize 297
244 1.8. Miscellaneous articles, coordinate conjunctions, or prepositions (unless the 298
245 Compare the following: title begins with such a word). Leave two blank lines after 299
246 $conf_a$ confa the title. 300
247 $\mathit{conf}_a$ conf a AUTHOR NAME(s) and AFFILIATION(s) are to be 301
248 See The TEXbook, p165. centered beneath the title and printed in Times 12-point, 302
249 The space after e.g., meaning “for example”, should not non-boldface type. This information is to be followed by 303
250 be a sentence-ending space. So e.g. is correct, e.g. is not. two blank lines. 304
251 The provided \eg macro takes care of this. The ABSTRACT and MAIN TEXT are to be in a two- 305
252 When citing a multi-author paper, you may save space column format. 306
253 by using “et alia”, shortened to “et al.” (not “et. al.” as “et” MAIN TEXT. Type main text in 10-point Times, single- 307
254 is a complete word). If you use the \etal macro provided, spaced. Do NOT use double-spacing. All paragraphs 308
255 then you need not worry about double periods when used at should be indented 1 pica (approx. 16 inch or 0.422 cm). 309
256 the end of a sentence as in Alpher et al. However, use it only Make sure your text is fully justified—that is, flush left and 310
257 when there are three or more authors. Thus, the following is flush right. Please do not place any additional blank lines 311
258 correct: “Frobnication has been trendy lately. It was intro- between paragraphs. 312
259 duced by Alpher [?], and subsequently developed by Alpher Figure and table captions should be 9-point Roman type 313
260 and Fotheringham-Smythe [?], and Alpher et al. [?].” as in ????. Short captions should be centred. 314
261 This is incorrect: “... subsequently developed by Alpher Callouts should be 9-point Helvetica, non-boldface type. 315
262 et al. [?] ...” because reference [?] has just two authors. Initially capitalize only the first word of section titles and 316
263 first-, second-, and third-order headings. 317
264
2. Formatting your paper FIRST-ORDER HEADINGS. (For example, 1. Intro- 318
265 duction) should be Times 12-point boldface, initially cap- 319
266 All text must be in a two-column format. The total allow- italized, flush left, with one blank line before, and one blank 320
267 able size of the text area is 6 78 inches (17.46 cm) wide by 8 87 line after. 321
268 inches (22.54 cm) high. Columns are to be 3 14 inches (8.25 SECOND-ORDER HEADINGS. (For example, 1.1. 322
5
269 cm) wide, with a 16 inch (0.8 cm) space between them. The Database elements) should be Times 11-point boldface, 323

3
CVPR CVPR
#***** #*****
CVPR 2023 Submission #*****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE.

324 378
325 379
326 380
327 381
328 382
329 383
330 384
331 385
332 386
333 387
334 388
335 389
336 390
337 (a) An example of a subfigure. (b) Another example of a subfigure. 391
338 392
339 Figure 2. Example of a short caption, which should be centered. 393
340 394
341 395
initially capitalized, flush left, with one blank line before, Method Frobnability
342 396
and one after. If you require a third-order heading (we dis-
343 Theirs Frumpy 397
courage it), use 10-point Times, boldface, initially capital-
344 Yours Frobbly 398
ized, flush left, preceded by one blank line, followed by a
345 Ours Makes one’s heart Frob 399
period and your text on the same line.
346 400
347 2.3. Footnotes Table 1. Results. Ours is better. 401
348 402
349 Please use footnotes1 sparingly. Indeed, try to avoid 403
350 footnotes altogether and include necessary peripheral ob- 2.5. References 404
351 servations in the text (within parentheses, if you prefer, as 405
List and number all bibliographical references in 9-point
352 in this sentence). If you wish to use a footnote, place it at the 406
Times, single-spaced, at the end of your paper. When ref-
353 bottom of the column on the page on which it is referenced. 407
erenced in the text, enclose the citation number in square
354 Use Times 8-point type, single-spaced. 408
brackets, for example [?]. Where appropriate, include page
355 numbers and the name(s) of editors of referenced books. 409
2.4. Cross-references
356 When you cite multiple papers at once, please make sure 410
357 For the benefit of author(s) and readers, please use the that you cite them in numerical order like this [?, ?, ?, ?, ?]. 411
358 If you use the template as advised, this will be taken care of 412
359
\cref{...} 413
automatically.
360 414
command for cross-referencing to figures, tables, equa-
361
tions, or sections. This will automatically insert the appro-
2.6. Illustrations, graphs, and photographs 415
362 416
priate label alongside the cross-reference as in this example: All graphics should be centered. In LATEX, avoid using
363 the center environment for this purpose, as this adds po- 417
364 To see how our method outperforms previous tentially unwanted whitespace. Instead use 418
365 work, please see ?? and ??. It is also possible to 419
366 \centering 420
refer to multiple targets as once, e.g. to ????. You
367 may also return to ?? or look at ??. at the beginning of your figure. Please ensure that any 421
368 point you wish to make is resolvable in a printed copy of the 422
369 If you do not wish to abbreviate the label, for example at the 423
beginning of the sentence, you can use the paper. Resize fonts in figures to match the font in the body
370 text, and choose line widths that render effectively in print. 424
371 Readers (and reviewers), even of an electronic copy, may 425
\Cref{...}
372 choose to print your paper in order to read it. You cannot 426
373 command. Here is an example: insist that they do otherwise, and therefore must not assume 427
374 that they can zoom in to see tiny details on a graphic. 428
375 ?? is also quite important. When placing figures in LATEX, it’s almost always best to 429
376 1 This is what a footnote looks like. It often distracts the reader from use \includegraphics, and to specify the figure width 430
377 the main flow of the argument. as a multiple of the line width as in the example below 431

4
CVPR CVPR
#***** #*****
CVPR 2023 Submission #*****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE.

432 486
\usepackage{graphicx} ...
433 \includegraphics[width=0.8\linewidth] 487
434 {myfile.pdf} 488
435 489
436 2.7. Color 490
437 491
Please refer to the author guidelines on the CVPR 2023
438 492
web page for a discussion of the use of color in your docu-
439 493
ment.
440 494
If you use color in your plots, please keep in mind that a
441 495
significant subset of reviewers and readers may have a color
442 496
vision deficiency; red-green blindness is the most frequent
443 497
kind. Hence avoid relying only on color as the discrimina-
444 498
tive feature in plots (such as red vs. green lines), but add a
445 499
second discriminative feature to ease disambiguation.
446 500
447 501
3. Final copy
448 502
449 You must include your signed IEEE copyright release 503
450 form when you submit your finished paper. We MUST have 504
451 this form before your paper can be published in the proceed- 505
452 ings. 506
453 Please direct any questions to the production editor in 507
454 charge of these proceedings at the IEEE Computer Soci- 508
455 ety Press: https://www.computer.org/about/ 509
456 contact. 510
457 511
458 512
459 513
460 514
461 515
462 516
463 517
464 518
465 519
466 520
467 521
468 522
469 523
470 524
471 525
472 526
473 527
474 528
475 529
476 530
477 531
478 532
479 533
480 534
481 535
482 536
483 537
484 538
485 539