1<!DOCTYPE html> 2<html> 3<head> 4 <meta charset="utf-8"> 5 <meta name="description" 6 content="Cross-device Collaborative Dataset for Localization"> 7 <meta name="keywords" content="visual localization, augmented-reality, human-robot interaction"> 8 <meta name="viewport" content="width=device-width, initial-scale=1"> 9 <title>CroCoDL: Cross-device Collaborative Dataset for Localization</title> 10 11 <!-- Global site tag (gtag.js) - Google Analytics -->
vendor: 66 bytes, lines 11-12
11 12 <script async src="https://www.googletagmanager.com/gtag/js?id=
12G-PYVRSFMDRL
vendor: 14 bytes, lines 12-13
12"></script> 13
13<script> 14
vendor: 155 bytes, lines 14-22
14window.dataLayer = window.dataLayer || []; 15 16 function gtag() { 17 dataLayer.push(arguments); 18 } 19 20 gtag('js', new Date()); 21 22 gtag('config', '
22G-PYVRSFMDRL
vendor: 6 bytes, lines 22-23
22'); 23
23</script>
23 24 25
25<script> 26 var slidePosition = 1; 27 SlideShow(slidePosition); 28 29 // forward/Back controls 30 function plusSlides(n) { 31 SlideShow(slidePosition += n); 32 } 33 34 // images controls 35 function currentSlide(n) { 36 SlideShow(slidePosition = n); 37 } 38 39 function SlideShow(n) { 40 var i; 41 var slides = document.getElementsByClassName("Containers"); 42 var circles = document.getElementsByClassName("dots"); 43 if (n > slides.length) {slidePosition = 1} 44 if (n < 1) {slidePosition = slides.length} 45 for (i = 0; i < slides.length; i++) { 46 slides[i].style.display = "none"; 47 } 48 for (i = 0; i < circles.length; i++) { 49 circles[i].className = circles[i].className.replace(" enable", ""); 50 } 51 slides[slidePosition-1].style.display = "block"; 52 circles[slidePosition-1].className += " enable"; 53 } 54 </script>
54 55 56<style> 57 .line-between { 58 border-bottom: 3px solid black; /* Adds a line between this row and the next */ 59 } 60</style> 61 62 <link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" 63 rel="stylesheet"> 64 65 <link rel="stylesheet" href="../static/css/bulma.min.css"> 66 <link rel="stylesheet" href="../static/css/bulma-carousel.min.css"> 67 <link rel="stylesheet" href="../static/css/bulma-slider.min.css"> 68 <link rel="stylesheet" href="../static/css/fontawesome.all.min.css"> 69 <link rel="stylesheet" 70 href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css"> 71 <link rel="stylesheet" href="../static/css/index.css"> 72 <link rel="icon" href="../static/images/favicon.svg"> 73 74
74<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
74 75
75<script defer src="../static/js/fontawesome.all.min.js"></script>
75 76
76<script src="../static/js/bulma-carousel.min.js"></script>
76 77
77<script src="../static/js/bulma-slider.min.js"></script>
77 78
78<script src="../static/js/index.js"></script>
78 79</head> 80<body> 81 82<nav class="navbar" role="navigation" aria-label="main navigation"> 83 <div class="navbar-brand"> 84 <a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false"> 85 <span aria-hidden="true"></span> 86 <span aria-hidden="true"></span> 87 <span aria-hidden="true"></span> 88 </a> 89 </div> 90 <div class="navbar-menu"> 91 <div class="navbar-start" style="flex-grow: 1; justify-content: center;"> 92 <a class="navbar-item" href="https://localizoo.com/"> 93 <span class="icon"> 94 <i class="fas fa-home"></i> 95 </span> 96 </a> 97 98 <div class="navbar-item has-dropdown is-hoverable"> 99 <a class="navbar-link"> 100 More Research 101 </a> 102 <div class="navbar-dropdown"> 103 <a class="navbar-item" href="/crocodl/"> 104 CroCoDL 105 </a> 106 <a class="navbar-item" href="https://lamar.ethz.ch/"> 107 LaMAR 108 </a> 109 <a class="navbar-item" href="/workshop/"> 110 ICCV 2025 Workshop - CroCoDL 111 </a> 112 </div> 113 </div> 114 </div> 115 116 </div> 117</nav> 118 119<center><img src="../static/images/CroCo1.png" width="1000"></center> 120<section class="hero"> 121 <div class="hero-body"> 122 <div class="container is-max-desktop"> 123 <div class="columns is-centered"> 124 <div class="column has-text-centered"> 125 <h1 class="title is-1 publication-title">Cross-device Collaborative Dataset for Localization</h1> 126 <div class="is-size-5 publication-authors"> 127 <span class="author-block"> 128 <a href="https://hermannblum.net/">Hermann Blum</a><sup>1,2</sup>,</span> 129 <span class="author-block"> 130 <a href="https://www.linkedin.com/in/alemercurio/?trk=public_post_comment_actor-image&originalSubdomain=ch">Alessandro Mercurio</a><sup>2</sup>,</span> 131 <span class="author-block"> 132 <a href="https://joshuaoreilly.com/">Joshua O'Reilly</a><sup>2</sup>, 133 </span> 134 <span class="author-block"> 135 <a href="https://www.linkedin.com/in/timengelbracht/?locale=de_DE">Tim Engelbracht</a><sup>2</sup>, 136 </span> 137 <span class="author-block"> 138 <a href="https://dusmanu.com/">Mihai Dusmanu</a><sup>3</sup>, 139 </span> 140 <span class="author-block"> 141 <a href="https://people.inf.ethz.ch/marc.pollefeys/">Marc Pollefeys</a><sup>2,3</sup>, 142 </span> 143 <span class="author-block"> 144 <a href="https://zuriabauer.com/">Zuria Bauer</a><sup>2</sup> 145 </span> 146 </div> 147 148 <div class="is-size-5 publication-authors"> 149 <span class="author-block"><sup>1</sup>Lamarr Institute/Uni Bonn</span> 150 <span class="author-block"><sup>2</sup>ETH Zurich</span> 151 <span class="author-block"><sup>3</sup>Microsoft</span> 152 </div> 153 154 <div class="column has-text-centered"> 155 <div class="publication-links"> 156 <!-- PDF Link. --> 157 <span class="link-block"> 158 <a href="https://openaccess.thecvf.com/content/CVPR2025/html/Blum_CroCoDL_Cross-device_Collaborative_Dataset_for_Localization_CVPR_2025_paper.html" 159 class="external-link button is-normal is-rounded is-dark"> 160 <span class="icon"> 161 <i class="fas fa-file-pdf"></i> 162 </span> 163 <span>Paper</span> 164 </a> 165 </span> 166 <span class="link-block"> 167 <a href="" 168 class="external-link button is-normal is-rounded is-dark"> 169 <span class="icon"> 170 <i class="ai ai-arxiv"></i> 171 </span> 172 <span>arXiv - Soon</span> 173 </a> 174 </span> 175 <!-- Video Link. --> 176 <!-- <span class="link-block"> 177 <a href="" 178 class="external-link button is-normal is-rounded is-dark"> 179 <span class="icon"> 180 <i class="fab fa-youtube"></i> 181 </span>
182 <span>Video - Soon</span> 183 </a> 184 </span> --> 185 <!-- Code Link. --> 186 <span class="link-block"> 187 <a href="https://github.com/cvg/crocodl-benchmark" 188 class="external-link button is-normal is-rounded is-dark"> 189 <span class="icon"> 190 <i class="fab fa-github"></i> 191 </span> 192 <span>Code</span> 193 </a> 194 </span> 195 <!-- Dataset Link. --> 196 <span class="link-block"> 197 <a href="https://huggingface.co/CroCoDL" 198 class="external-link button is-normal is-rounded is-dark"> 199 <span class="icon"> 200 <i class="far fa-images"></i> 201 </span> 202 <span>Data</span> 203 </a> 204 </span> 205 <span class="link-block"> 206 <a href="https://www.codabench.org/competitions/9471/" 207 class="external-link button is-normal is-rounded is-dark"> 208 <span class="icon"> 209 <i class="fas fa-trophy"></i> 210 </span> 211 <span>Challenge</span> 212 </a> 213 </span> 214 </div> 215 216 </div> 217 </div> 218 </div> 219 </div> 220 </div> 221</section> 222 223<section class="hero teaser"> 224 <div class="container is-max-desktop"> 225 <div class="hero-body"> 226 <img src="../static/images/CrocoTeaser.png" width="2000" height="1000"> 227 <h2 class="subtitle has-text-centered"> 228 <span class="dnerf">CroCoDL:</span> the first dataset to contain sensor 229 recordings from real-world robots, phones, and mixed-reality 230 headsets, covering a total of 10 challenging locations to 231 benchmark cross-device and human-robot visual registra- 232 tion. 233 </h2> 234 </div> 235 </div> 236</section> 237 238 239<section class="section"> 240 <div class="container is-max-desktop"> 241 <!-- Abstract. --> 242 <div class="columns is-centered has-text-centered"> 243 <div class="column is-four-fifths"> 244 <h2 class="title is-3"><span class="dnerf">Abstract</span></h2> 245 <div class="content has-text-justified"> 246 <p> 247 Accurate localization plays a pivotal role in the autonomy of systems operating in unfamiliar environments, particularly when interaction with humans is expected. High-accuracy visual localization systems encompass various components, such as feature extractors, matchers, and pose estimation methods. This complexity translates to the necessity of robust evaluation settings and pipelines. However, existing datasets and benchmarks primarily focus on single-agent scenarios, overlooking the critical issue of cross-device localization. Different agents with different sensors will show their own specific strengths and weaknesses, and the data they have available varies substantially. 248 </p> 249 <p> 250 This work addresses this gap by enhancing an existing augmented reality visual localization benchmark with data from legged robots, and evaluating human-robot, cross-device mapping and localization. Our contributions extend beyond device diversity and include high environment variability, spanning ten distinct locations ranging from disaster sites to art exhibitions. Each scene in our dataset features recordings from robot agents, hand-held and head-mounted devices, and high-accuracy ground truth LiDAR scanners, resulting in a comprehensive multi-agent dataset and benchmark. This work represents a significant advancement in the field of visual localization benchmarking, with key insights into the performance of cross-device localization methods across diverse settings. 251 </p> 252 </div> 253 </div> 254 </div> 255 <!--/ Abstract. --> 256 </div> 257</section> 258 259 260<section class="hero is-light is-small"> 261 <div class="hero-body"> 262 <div class="container"> 263 <h2 class="title is-3"><center><span class="dnerf">Renderings vs. real CroCoDL data</span></center></h2> 264 <div id="results-carousel" class="carousel results-carousel"> 265 <div class="item item-steve">
266 <img src="../static/images/armadillo/dataset1.png" alt="Steve", height="475"> 267 </div> 268 <div class="item item-steve"> 269 <img src="../static/images/armadillo/dataset2.png" alt="Steve", height="475"> 270 </div> 271 <div class="item item-steve"> 272 <img src="../static/images/armadillo/dataset3.jpg" alt="Steve", height="475"> 273 </div> 274 <div class="item item-steve"> 275 <img src="../static/images/armadillo/dataset6.jpg" alt="Full Body"> 276 </div> 277 <div class="item item-steve"> 278 <img src="../static/images/armadillo/dataset8.jpg" alt="Blue Shirt"> 279 </div> 280 <div class="item item-steve"> 281 <img src="../static/images/armadillo/dataset10.png" alt="Mask"> 282 </div> 283 <div class="item item-steve"> 284 <img src="../static/images/armadillo/dataset13.png" alt="Coffee"> 285 </div> 286 <div class="item item-steve"> 287 <img src="../static/images/armadillo/dataset17.png" alt="Toby"> 288 </div> 289 <div class="item item-steve"> 290 <img src="../static/images/armadillo/dataset19.png" alt="Toby"> 291 </div> 292 <div class="item item-steve"> 293 <img src="../static/images/armadillo/dataset20.png" alt="Toby"> 294 </div> 295 <div class="item item-steve"> 296 <img src="../static/images/armadillo/dataset22.png" alt="Toby"> 297 </div> 298 <div class="item item-steve"> 299 <img src="../static/images/armadillo/dataset27.png" alt="Toby"> 300 </div> 301 </div> 302 <br> 303 <h2 class="subtitle has-text-centered"> 304 <span class="dnerf">Qualitative Results.</span> Good alignment between rendering and real 305 image validates correct camera pose with respect to the NavVis-based ground truth scan from which we render here. 306 </h2> 307 </div> 308 </div> 309</section> 310 311<section class="hero teaser"> 312 <div class="container is-max-desktop"> 313 <div class="hero-body"> 314 <br> 315 <h2 class="title is-3"><center><span class="dnerf">Locations</span></center></h2> 316 <img src="../static/images/Locations.png" width="2500" height="1000"> 317 <center> 318 <img src="../static/images/Locations2.png" width="800" height="1000"></center> 319 <h2 class="subtitle has-text-centered"> 320 <span class="dnerf">New locations of the CroCoDL dataset.</span> Each location has high-quality meshes, obtained from LiDAR, which 321 are registered with numerous phone, AR headset, and robotic sequences. 322 </h2> 323 </div> 324 </div> 325</section> 326 327<section class="hero is-light is-small"> 328 <div class="hero-body"> 329 <div class="container"> 330 <h2 class="title is-3"><span class="dnerf"><center>Commonly used datasets for visual localization and SLAM</center></span></h2> 331<table border="0" style="border-collapse: collapse; text-align:center; width: 100%;"> 332 <br> 333 <thead> 334 <tr> 335 <th>Dataset</th> 336 <th>Motion</th> 337 <th>Env.</th> 338 <th>Locations</th> 339 <th>Changes</th> 340 <th>Sensors</th> 341 <th>GT pose accuracy</th> 342 <th>Seqs.</th> 343 </tr> 344 </thead> 345 <tbody> 346 <tr> 347 <td>KITTI</td> 348 <td>ð</td> 349 <td>â¬</td> 350 <td>1</td> 351 <td>ðð¦ï¸</td> 352 <td>RGB, LiDAR, IMU</td> 353 <td><10cm (RTKGPS)</td> 354 <td>22</td> 355 </tr> 356 <tr> 357 <td>TUM RGBD</td> 358 <td>âð¼</td> 359 <td>â¬</td> 360 <td>2</td> 361 <td>ð</td> 362 <td>RGB-D, IMU</td> 363 <td>1mm (mocap)</td> 364 <td>80</td> 365 </tr> 366 <tr> 367 <td>Malaga</td> 368 <td>ð</td> 369 <td>â¬</td> 370 <td>1</td> 371 <td>ðð¦ï¸</td> 372 <td>RGB, IMU</td> 373 <td>(GPS)</td> 374 <td>15</td> 375 </tr> 376 <tr> 377 <td>EUROC</td> 378 <td>ð¸</td> 379 <td>â¬</td> 380 <td>2</td> 381 <td>-</td> 382 <td>RGB, IMU</td> 383 <td>1mm (mocap)</td> 384 <td>11</td> 385 </tr> 386 <tr> 387 <td>NCLT</td> 388 <td>ð¼</td> 389 <td>â¬â¬</td> 390 <td>1</td> 391 <td>ððªðð¦ï¸</td> 392 <td>RGB, LiDAR, IMU, GNSS</td> 393 <td><10cm (GPS + IMU + LiDAR)</td> 394 <td>27</td> 395 </tr> 396 <tr> 397 <td>PennCOSYVIO</td> 398 <td>â</td> 399 <td>â¬â¬</td> 400 <td>1</td> 401 <td>ðð¦ï¸</td> 402 <td>RGB, IMU</td> 403 <td>15cm (visual tags)</td> 404 <td>4</td> 405 </tr> 406 <tr> 407 <td>TUM VIO</td> 408 <td>â</td> 409 <td>â¬â¬</td> 410 <td>4</td> 411 <td>-</td> 412 <td>RGB, IMU</td> 413 <td>1mm (mocap ends)</td> 414 <td>28</td> 415 </tr> 416 <tr> 417 <td>
417UZH-FPV</td> 418 <td>ð¸</td> 419 <td>â¬â¬</td> 420 <td>2</td> 421 <td>-</td> 422 <td>RGB, event camera, IMU</td> 423 <td>~1cm (total station + VI-BA)</td> 424 <td>28</td> 425 </tr> 426 <tr> 427 <td>ETH3D SLAM</td> 428 <td>â</td> 429 <td>â¬</td> 430 <td>1</td> 431 <td>-</td> 432 <td>RGB, depth, IMU</td> 433 <td>1mm (mocap)</td> 434 <td>96</td> 435 </tr> 436 <tr> 437 <td>Newer College</td> 438 <td>â</td> 439 <td>â¬â¬</td> 440 <td>1</td> 441 <td>-</td> 442 <td>RGB, LiDAR, IMU</td> 443 <td>3cm (LiDAR ICP)</td> 444 <td>3</td> 445 </tr> 446 <tr> 447 <td>OpenLoris-Scene</td> 448 <td>ð¼</td> 449 <td>â¬</td> 450 <td>5</td> 451 <td>ððª</td> 452 <td>RGB-D, IMU, wheel odom.</td> 453 <td><10cm (2D LiDAR)</td> 454 <td>22</td> 455 </tr> 456 <tr> 457 <td>TartanAir</td> 458 <td>syn.</td> 459 <td>â¬â¬</td> 460 <td>30</td> 461 <td>-</td> 462 <td>RGB</td> 463 <td>perfect (synthetic)</td> 464 <td>30</td> 465 </tr> 466 <tr> 467 <td>UMA VI</td> 468 <td>âð</td> 469 <td>â¬â¬</td> 470 <td>2</td> 471 <td>-</td> 472 <td>RGB, IMU</td> 473 <td>(visual tags)</td> 474 <td>32</td> 475 </tr> 476 <tr> 477 <td>UrbanLoco</td> 478 <td>ð</td> 479 <td>â¬</td> 480 <td>12</td> 481 <td>ððª</td> 482 <td>RGB, LiDAR, IMU, GNSS, SPAN-CPT</td> 483 <td>12</td> 484 </tr> 485 <tr> 486 <td>Naver Labs</td> 487 <td>ð¼</td> 488 <td>â¬</td> 489 <td>5</td> 490 <td>ð ðª</td> 491 <td>RGB, LiDAR, IMU</td> 492 <td><10cm (LiDAR SLAM <br>and SfM)</td> 493 <td>10</td> 494 </tr> 495 <tr> 496 <td>HILTI SLAM</td> 497 <td>â</td> 498 <td>â¬â¬</td> 499 <td>8</td> 500 <td>-</td> 501 <td>RGB, LiDAR, IMU</td> 502 <td><5mm (total station)</td> 503 <td>12</td> 504 </tr> 505 <tr> 506 <td>Graco</td> 507 <td>ð¼ð¸</td> 508 <td>â¬</td> 509 <td>1</td> 510 <td>ð ð¦ï¸</td> 511 <td>RGB, LiDAR, GPS, IMU</td> 512 <td>â1cm (GNSS)</td> 513 <td>14</td> 514 </tr> 515 <tr> 516 <td>FusionPortable</td> 517 <td>2 ∈ âð¦¿ð¼ð</td> 518 <td>â¬â¬</td> 519 <td>9</td> 520 <td>-</td> 521 <td>RGB, event cameras, LiDAR, IMU, GPS</td> 522 <td>â1cm (GNSS RTK)</td> 523 <td>41</td> 524 </tr> 525 <tr class="line-between"> 526 <td>LaMAR</td> 527 <td>âð¥½</td> 528 <td>â¬â¬</td> 529 <td>3</td> 530 <td>ððªððï¸ð¦ï¸</td> 531 <td>RGB, LiDAR, depth, IMU, WiFi/BT</td> 532 <td><10cm (LiDAR + PGO + PGO-BA)</td> 533 <td>500</td> 534 </tr> 535 <tr> 536 <td><b>CroCoDL</b></td> 537 <td>âð¥½ð¦¿[ð¸]</td> 538 <td>â¬â¬</td> 539 <td>10</td> 540 <td>ððªððï¸ð¦ï¸</td> 541 <td>RGB, LiDAR, depth, IMU, WiFi/BT</td> 542 <td>~10cm (LiDAR + PGO + PGO-BA)</td> 543 <td>500 <strong>+800</strong></td> 544 </tr> 545 </tbody> 546 </table> 547 <br> 548 <center><h3 class="title is-3"><span class="dnerf"><p><strong>Legend</strong> </p></h3></span> 549 <p><strong><span class="dnerf">Environment:</span></strong><span style="background-color: #f0f0f0; padding: 2px 5px; border-radius: 4px;">⬠inside</span>, 550 <span style="background-color: #f0f0f0; padding: 2px 5px; border-radius: 4px;">⬠outside</span>;</p> 551 552 <p><strong><span class="dnerf">Changes:</span></strong> ð Structural changes due to moving people, ðª long-term changes due to displaced furniture, ð¦ï¸ weather, ð day-night, ðï¸ construction work;</p> 553 554 <p><strong><span class="dnerf">Trajectory motion from sensors mounted on:</span></strong> ð¼ ground vehicle, 𦿠legged robot, ð¸ drone, ð car, â hand-held, 𥽠head-mounted, <em>âsyn.â synthetic.</em></p> 555 556 <p>(noted with *: at most 2 devices are recorded in the same location; [ð¸]: not aligned, due to safety / permission reasons - we could only capture drone footage in 8/10 locations)</p> 557 </center> 558 </div> 559 </div> 560</section> 561<br> 562<section class="section" id="BibTeX">
563 <div class="container is-max-desktop content"> 564 <h2 class="title">BibTeX</h2> 565 <pre><code>@inproceedings{blum2025crocodl, 566 author = {Blum, Hermann and Mercurio, Alessandro and O'Reilly, Josua and Engelbracht, Tim and Dusmanu, Mihai and Pollefeys, Marc and Bauer, Zuria}, 567 title = {CroCoDL: Cross-device Collaborative Dataset for Localization}, 568 booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition}, 569 year = {2025}, 570}</code></pre> 571 </div> 572 <br> 573 <center> 574 <h2 class="title">CroCoDL Team</h2> 575 <img src="../static/images/cvg_logo_horizontal-1.svg" width="450" height="1000"> 576 <img src="../static/images/logo_text.svg" width="450" height="900" style="position: relative; top: 21px;"> 577 <br> 578 <br> 579 <img src="../static/images/eth_logo_kurz_pos.svg" width="300" height="1000" style="position: relative; top: 10px;"> 580 <img src="../static/images/UNI_Bonn_Logo_Kompakt.jpg" width="150" height="700" style="position: relative; top: 21px;"> 581 582 <img src="../static/images/Microsoft_Logo.png" width="250" height="900" style="position: relative; top: -21px;"> 583</center> 584</section> 585 586 587<footer class="footer"> 588 <div class="container"> 589 <div class="content has-text-centered"> 590 591 </div> 592 <div class="columns is-centered"> 593 <div class="column is-8"> 594 <div class="content"> 595 <p> 596 This website uses source code originally from <a 597 href="https://github.com/nerfies/nerfies.github.io">Nerfies</a>, licensed under a <a rel="license" 598 href="http://creativecommons.org/licenses/by-sa/4.0/">Creative 599 Commons Attribution-ShareAlike 4.0 International License</a>. 600 </p> 601 </div> 602 </div> 603 </div> 604 </div> 605</footer> 606 607</body> 608</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.